JP2643295B2 - Fuzzy control arithmetic unit and fuzzy inference method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an input and / or output of a control measurement value and / or a quantitative value for sensuously determining the control amount. The present invention relates to a fuzzy control operation device that performs an operation according to a fuzzy inference rule according to an inference rule, creates an output membership function as a result, determines an output value from the function, and outputs it as a control operation amount.

2. Description of the Related Art Fuzzy means "ambiguity" in Japanese. Fuzzy theory is a set theory proposed by Professor LAZadeh of the United States in 1965, but fuzzy inference related to the present invention is a field of fuzzy theory. (For example, pages 873 to 880 of the Society of Instrument and Control Engineers, 20 (1983),
Systems and Controls 28 (1984) pp. 597-604).

Fuzzy control is said to be comparable to conventional PID control and modern control, especially if the control system does not handle nonlinear system systems or ambiguous information that involves human evaluation. In a system that is not good, the knowledge and control experience of a skilled person can be handled, and its effectiveness is high.

Therefore, it is considered that the present invention is effective not only for a linear control output such as a PID control device widely used in a process control system but also for a control target requiring an arbitrary non-linear control output. In other words, in a complex control system system in which a mathematical model cannot be constructed using fuzzy control, the control that realizes the system is reproduced by a computer based on the knowledge and control experience that humans (operators) have gained from conventional experience. I'm in the place to try.

In addition, there are papers and patents that use the terms fuzzy inference and fuzzy control for the terms fuzzy inference and fuzzy control. However, in this specification, all terms will be unified to the term “fuzzy”. did.

Before explaining the conventional technique, first, a basic outline of fuzzy inference (also referred to as fuzzy control) will be described.

In a control system that involves human evaluation, variables that are subjectively and intuitively determined by the operator as control variables, for example, variables such as “large”, “medium”, “very”, and “slightly”. (This is called a fuzzy variable) in some cases to determine the final control operation amount. This input variable determines the operation amount using the control experience obtained from the operator's own experience. The fuzzy control arithmetic unit applies the fuzzy variables of the input variables to the antecedent part (IF part) of the inference rule according to a predetermined inference rule, and determines the consequent part (THEN part) from the satisfaction level (membership value) for the rule. Determines the fuzzy number of the output. Here, the fuzzy number means an ambiguous numerical value such as “approximately 2” and is expressed by a membership function. The actual operation amount can be obtained by taking the center of gravity value of the fuzzy number of the output according to a plurality of rules.

FIG. 9 is a block diagram showing a conventional example of a fuzzy control arithmetic unit. Each block will be described as follows.
In the figure, 1 is a data (including a calculated value and a human evaluation value) input device, 2 is a display command device, 3 is a fuzzy inference operation device, 4 is an inference result output device, and 5 is a display device.

The display command device 2 is constituted by a keyboard, and the fuzzy inference operation device 3 is constituted by a digital computer. The data input to the data input device 1 is subjected to an inference operation by the fuzzy inference operation device 3 and outputs an inference result. At the same time, a list of inference rules, a list of fuzzy variables, and use of various inference rules are displayed on the display device 5. The status etc. are displayed.

In FIG. 9 and FIG. 9, the inference rules of fuzzy inference rules and fuzzy variables as input data are fixed at a fuzzy inference operation device and have no function of changing fuzzy variables.

FIG. 10 is a block diagram showing another example of a conventional fuzzy control operation device capable of tuning an inference rule. Tenth
The figure is different from FIG. 9 in that a data input device 1 for inputting a fuzzy number, a fuzzy inference operation device 3 for performing a fuzzy inference operation using the input value, and a control based on the fuzzy number of the inference result obtained as a result. An inference result device 4 that outputs a control output as a result, and a display device 5 that displays input values, fuzzy inference rules and fuzzy variables used in fuzzy inference, or inference result output, etc., according to a command from the display command device 2. The parameters input from the inference abnormality detection parameter input device 109 and the fuzzy number of the inference result obtained by the fuzzy inference to the fuzzy inference device of FIG. 9 are given from the fuzzy inference operation device 103, and are compared with the above parameters. From the examination of the inference results, the fuzzy inference operation unit 103 detects the inference abnormality and the inference abnormality detection unit 110 and the inference abnormality When There sensed, the inference abnormal state storage device 111 for storing various related data at that time as inference condition data, inferring the abnormal situation display instruction unit 112
Abnormal status display device 113 for displaying the date and time of the occurrence of the abnormality related to the inference abnormality, the input value at that time, the inference rule, the output value, the fuzzy number of the inference result, etc.
Is added.

However, in such a conventional method, the variables in the consequent part of fuzzy inference are given by fuzzy variables.
It is difficult to predict how the nonlinear system output can be obtained if the rules are described, and tuning of the inference rules is difficult because of the large number of parameters to be tuned. In many cases, the simplicity and high speed of linear system systems, which are the advantages of conventional PID control, could not be included.

Therefore, the present invention provides a fuzzy control operation device that can reduce the storage capacity of a storage unit for storing fuzzy inference rules, facilitate rule adjustment, and speed up fuzzy inference operation. Aim.

Means for Solving the Problems In order to achieve the above object, a fuzzy control arithmetic device according to the present invention inputs data of both or one of an observation value from a controlled object and a quantitative value determined by a human being sensuously. An operation device for performing a fuzzy inference operation to output an operation amount to a control target, comprising: a data input unit for inputting the data; and a fuzzy storage unit for storing an inference rule with a consequent part of the fuzzy inference rule being a constant value. An inference rule storage unit, performing each of the fuzzy inference operations to match the antecedent part of the plurality of fuzzy inference rules with respect to the input data, and a constant value of a consequent part of the stored fuzzy inference rule. A fuzzy inference operation unit that obtains the output value of each fuzzy inference rule from And an output unit for outputting the obtained fuzzy inference operation result.

Further, the fuzzy inference method of the present invention is a fuzzy inference method for inferring an operation amount to the control object from data of both or any of the observation value of the control object and the quantitative value determined by human being sensuously. , Including the following steps.

Storing a constant value in a memory in advance as a consequent part of the fuzzy inference rule; inputting the data;
Obtaining respective degrees of conformity to the antecedent part of the plurality of fuzzy inference rules with respect to the input data; and fuzzy inference from the respective degrees of conformity and the constant value of the consequent part of the fuzzy inference rule stored in advance. A step of obtaining an output value of a rule; and a step of obtaining an operation amount for a control target from an output value of each fuzzy inference rule.

Function As described above, by setting the consequent part of the fuzzy inference rule to a constant value, a small-capacity memory can be used for storing the fuzzy inference rule, facilitating rule adjustment and speeding up fuzzy inference operation. Becomes

Embodiment An embodiment of the present invention will be described below with reference to the drawings.

In FIG. 1, reference numeral 11 denotes a control status input unit for inputting measurement status in process control, automatic adjustment control, and servo mechanism control. Reference numeral 12 denotes a language evaluation input unit, which is also an input unit, and inputs a control evaluation value in the current situation of the operator.
For example, an ambiguous evaluation amount such as "good", "very good", or "slightly good" is input using words. Reference numeral 14 denotes a fuzzy number storage unit which stores the fuzzy numbers of the above-mentioned ambiguous evaluation amounts, such as "negative large", "zero", "positive small" used in the antecedent part of the fuzzy inference operation. A fuzzy variable representing the magnitude of the control amount is also stored.

13 is a fuzzy number input change section, which is the language evaluation input section described above.
The fuzzy number corresponding to the evaluation expression represented by 12 is extracted from the fuzzy number storage unit 14 and changed as an input. For example, when an ambiguous evaluation amount based on the word "good" is input to the language evaluation input unit 12, the fuzzy number corresponding to "good" is stored in the fuzzy number storage unit 14 as a form indicated by the membership function in FIG. Fuzzy number input change section
In 13, the word "good" is changed to a fuzzy number represented by a membership function. Note that fuzzy information is used as a language evaluation input.

Reference numeral 15 denotes a fuzzy inference rule storage unit, which stores many inference rules used for fuzzy inference. The rules used here are expressed in the form "IF ... (antecedent) ... THEN ... (consequent)", where the antecedent is an inference proposition containing fuzzy variables.
The consequent part is represented by a constant value that is a normal real number. Reference numeral 16 denotes an inference rule management unit which extracts a fuzzy number from the fuzzy number storage unit 14 and an inference rule from the fuzzy inference rule storage unit 15.

Reference numeral 17 denotes a fuzzy inference operation unit which applies a fuzzy number, which is an output of the control status input unit 11 and a fuzzy number, which is an output of the fuzzy number input change unit 13, to a fuzzy inference and derives an inference result. . Reference numeral 18 denotes a control manipulated variable output unit that outputs a result of fuzzy inference. 19 is a control command display section,
Displays control commands. Reference numeral 20 denotes a control status display unit, which displays the status of the fuzzy inference and the result according to a command from the control command display unit 19. The above is the basic device for performing the fuzzy control calculation.

The fuzzy control arithmetic device configured as described above and a fuzzy inference method that is easy to tune will be described in detail below.

<Inference method> In the conventional fuzzy control, information obtained from the control input, for example, considering a first-order lag tank model, the relationship between the control error e and its change rate de and the control output (manipulated variable) u
If described as IF ... THEN rule, prepare multiple inference rules as follows: IF e is Zero and de is Positive Medium THEN u is Negative Medium. IF… part as antecedent part, THEN
The part of ... is called a consequent part. Zero, Positive Medium, Negative Medium, etc. are labels indicating the fuzzy numbers of inputs and outputs used for describing rules, and are fuzzy variables. FIG. 2 shows an example.

In the present invention, the fuzzy variables in the antecedent part are Negative Big (NB), Negative M
edium (NM), Negative Small (NS), Zero (ZO), Positi
ve Small (PS), Positive Medium (PM), Positive Big
Each fuzzy variable in (PB) is set to have a fixed distance. Also, since the actual control output is usually a normal real value, u is not a fuzzy number but a real value. That is, IF e is Zero and de is Positive Medium THEN u = f (for example, real value 2). Becomes FIG. 3 shows an embodiment of the fuzzy inference process.
Rule tables for the normal fuzzy control and the proposed fuzzy control are shown in FIGS. 4 and 5, respectively.

The fuzzy number of the conclusion from the above rule is Which indicates the degree of conformity of the antecedent part. Since there are a plurality of inference rules, the combined result T is obtained as a fuzzy number of u: μ _T (u) = μ ₁ (u) Vμ ₂ (u) V... Vμ _n (u) (2) Here, Λ and V represent a min operation and a max operation, respectively.

As a method of converting the fuzzy number T into a manipulated variable (real value),
When weighted center of gravity is adopted, Becomes Where f is the real value of the consequent.

The output of the PID control and the output of the proposed fuzzy control are shown in FIGS. 6 and 7, respectively. FIG. 6 shows an output example of a PD controller composed of two elements, a proportional element and a differential element (the integral element is omitted for display), and the control output u corresponding to e and de is plotted on the vertical axis.

When the inference rule is given as shown in FIG. 5 in the proposed fuzzy control, the same output as in the PD control in FIG. 6 can be obtained. Further, by adding an integral element to the input of the fuzzy control, the same output as that of the PID control can be obtained in the same manner. When the rule is changed as shown in FIG. 8 to have a nonlinear component, the output becomes as shown in FIG.

When e is NB and de is PB, the output is lowered, and e is PB and de is
The output is increased near NB. In this way, it is easy to give the PID control a non-linear component based on the rule table of FIG. The features of the proposed fuzzy control are quadratic, 3
It is considered that the desired nonlinear output can be obtained much more easily than using a higher-order function such as

In the present embodiment, the weight of the center of gravity of the fuzzy number T is used as the control operation amount obtained as a result of the fuzzy inference.

In the present embodiment, the min and max operations are used in equations (5) and (2) of the fuzzy inference operation, respectively.
t-nor for algebraic, marginal, and intense products instead of n operations
The m-operation may be replaced by a t-conorm operation such as an algebraic sum, a limit sum, or an intense sum instead of the max operation.

As described above, according to the present invention, the consequent part of the fuzzy inference rule is set to a constant value, so that the storage capacity of the storage unit for storing the fuzzy inference rule can be reduced and the rule adjustment can be performed. It is possible to simplify and speed up the fuzzy inference operation.

Furthermore, if the constant values of the consequent are taken at equal intervals,
Fuzzy control is PID control, but even in this case, the inference rule of the nonlinear system can be easily tuned only by changing the real value of the consequent part, and control can be performed by fuzzy inference after tuning is completed. large.

[Brief description of the drawings]

FIG. 1 is a block diagram of a fuzzy control operation device of one embodiment of the present invention, FIG. 2 is a diagram showing fuzzy variables of one embodiment of the present invention, and FIG. FIG. 4 is a diagram showing an inference rule of a conventional fuzzy control, FIG. 5 is a diagram showing an inference rule of an example of fuzzy control in the apparatus, FIG. Is a diagram showing a control output of one embodiment by PID control, FIG. 7 is a diagram showing a control output of one embodiment of the fuzzy control proposed in the present invention, and FIG. 8 is a diagram showing inference rules in FIG. FIG. 9 is a configuration diagram of a conventional fuzzy control arithmetic device, and FIG. 10 is a configuration diagram of a conventional fuzzy control arithmetic device capable of tuning an inference rule. 11 ... control status input section, 12 ... language evaluation input section, 13 ...
Fuzzy number input change unit, 14 ... Fuzzy number storage unit, 15 ...
... Fuzzy inference rule storage unit, 16 ... Inference rule management unit, 17
…… Fuzzy inference operation part, 18 …… Control operation amount output part, 19
... Control command display section, 20 ... Control status display section.

 ────────────────────────────────────────────────── ─── Continuation of the front page (56) References Toshiro Terano and two others “Introduction to Fuzzy Systems” (Showa 62-4-30) Ohmsha Publishing Co., Ltd., pp. 153-155 Fuji Times, 58 [4] (Showa 60), Michio Sugano and three others "General-purpose fuzzy control system", 307-314 System and Control Tutorial Lecture '87 “Guidelines for Control Engineering (Latest Modeling Theory and Know-How)” Textbook, (Showa 62), Japan Automated Control Association, Michio Sugano, “Fuzzy Modeling of Systems,” . 79−98

Claims (2)

(57) [Claims] 1. An operation for inputting data or both of an observation value from a controlled object and a quantitative value determined by a human being sensuously, performing a fuzzy inference operation, and outputting an operation amount to the controlled object. An apparatus, comprising: a data input unit for inputting the data; a fuzzy inference rule storage unit for storing an inference rule with a consequent part of the fuzzy inference rule being a constant value; and the input data by performing the fuzzy inference operation. The output value of each fuzzy inference rule is obtained from the respective degrees that match the antecedent part of the plurality of fuzzy inference rules and the stored constant value of the consequent part of the fuzzy inference rule. A fuzzy inference operation unit for integrating the output values to obtain an operation amount for the controlled object; and an output unit for outputting the obtained fuzzy inference operation result. Fuzzy control arithmetic unit. 2. A fuzzy inference method for inferring an operation amount to a control target from data of at least one of an observation value of the control target and a quantitative value determined by a human being,
It includes the following steps. Storing a constant value in a memory in advance as a consequent part of the fuzzy inference rule; inputting the data; obtaining respective degrees of conformity with the antecedent part of the plurality of fuzzy inference rules for the input data; Obtaining the output value of each fuzzy inference rule from the degree of conformity of the fuzzy inference rule and the constant value of the consequent part of the fuzzy inference rule stored in advance, obtaining the operation amount to the control object from the output value of each fuzzy inference rule Step.