JPH05241612A - Fuzzy controller - Google Patents

Abstract

(57) [Abstract] [Purpose] The present invention is to efficiently execute control rules and membership function tuning. A fuzzy controller that executes a fuzzy control calculation using a control rule and a membership function and applies the manipulated variable obtained by this fuzzy control calculation to a controlled object 26 is fuzzy labeled by the membership function. A fitness display table 24C 'that stores the fitness of each fuzzy set of input variables corresponding to the controlled variable, and a human interface device 27 that inputs various information for tuning control rules and membership functions. , A fuzzy controller provided with display means 27 and 28 for performing tuning processing by reading out the goodness of fit of each fuzzy set stored in the goodness-of-fit display table and displaying it on the human interface device.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fuzzy controller used for fuzzy control calculation of various control systems, and more particularly to a fuzzy controller that efficiently executes control rules and membership function tuning.

[0002]

2. Description of the Related Art Recently, regarding control knowledge possessed by humans,
While represented by an inference rule (hereinafter referred to as a control rule) including a condition part (if part) and a conclusion part (then part),
After taking in the controlled variable (process amount) obtained from the controlled object, estimating the qualitative manipulated variable based on the control rule, and further obtaining the quantitative manipulated variable using the membership function, this manipulated variable is calculated. A fuzzy controller for applying to a controlled object has been proposed.

In this fuzzy controller, as shown in FIG. 4, a deviation calculating means 1 obtains a control deviation E from a controlled variable X from a controlled object 2 and a target value S thereof, and then the fuzzy calculating device calculates the control deviation E. 3, the fuzzy control operation is executed using the control rule and the membership function, which are preliminarily stored in the storage unit 4 and are composed of the condition part and the conclusion part, and the operation amount M as the operation result is used as the controlled object 2 It is configured to be applied to.

In particular, this fuzzy controller is effective for controlling a system having an uncertain ambiguity, and is therefore widely used not only in plants but also in various control fields.

Here, the membership function is a function that characterizes a fuzzy set, and corresponds to a characteristic function in a normal set. Generally, it is defined by the format shown in FIG. That is, the input value on the horizontal axis belongs to each fuzzy set, and the degree on the vertical axis (fitness)
It is a function group that shows as an output value. On the other hand, the control rule describes only the qualitative relationship between the input and output variables to be controlled, and is based on the operator's control knowledge and so on.

If Define in the then format. Among them, the part "if-if" is the antecedent part of the control rule described above, while the part "... Do" corresponds to the consequent part of the control rule.

When the antecedent part has two variables and the consequent part has one variable, this control rule is usually expressed in a table format as shown in FIG. 6, for example. In other words, this table is N
It is a rule table that gives fuzzy output to two human powers fuzzyly divided by 7 types of fuzzy labels consisting of B, NH, NS, ZO, PS, PM and PB.

[0008]

Therefore, in the fuzzy controller having the above-mentioned structure, when the controlled object is always controlled in the best condition, it is necessary that the control rule and the membership function are always set correctly. Moreover, if the control rules are configured during simulation or adjustment by the actual machine to the target plant. then
It is necessary to tune each rule and membership function such as peak value, tilt angle, fuzzy labeling (fuzzy division), etc., but this tuning work is very complicated. Is.

Generally, in a PID controller that executes a normal linear control operation that is not fuzzy control, a mathematical model is used to logically determine the constants and coefficients of each control operation expression, so the tuning operation is relatively short. Can be done.

However, since the setting of the control rules and membership functions in fuzzy control is probably dependent on the experience of skilled engineers, the shape of each control rule and each membership function is theoretically determined. It is not possible. Therefore, at the present time, engineers perform tuning work by trial and error. Therefore, much time and labor are required until the best control state for the controlled object is obtained.

As described above, the conventional fuzzy controller has many trial and error elements for tuning, and the design method based on theoretical support has not been established yet. There is a problem that tuning takes a very long time.

The present invention has been made in view of the above-mentioned circumstances, and stores the degree of conformity of fuzzy variables in a table in advance, which can be used as a guide for changing the type of membership function or reducing the number of fuzzy labels during simulation or adjustment by an actual machine. The purpose is to provide a fuzzy controller that can perform tuning efficiently by using it.

[0013]

In order to solve the above problems, the present invention executes a fuzzy control operation using a control rule and a membership function, and applies an operation amount obtained by this fuzzy control operation to a controlled object. In the fuzzy controller, the goodness-of-fit display table for storing the goodness of fit of each fuzzy set of the input variable corresponding to the control amount fuzzy labeled by the membership function, the control rule and the membership function The human interface device for inputting various information for tuning and the fitness of each fuzzy set stored in the fitness display table are read out and displayed on the human interface device to display the control rule and the membership function. Tuning tuning A fuzzy controller having a shown means.

[0014]

Therefore, according to the present invention, by taking the above-mentioned means, during the adjustment by the simulation or the actual machine,
The membership function is read by reading the fitness of each fuzzy set of the input variables corresponding to the control amount fuzzy labeled by the membership function stored in the fitness display table and displaying it on the Human interface device. It can be used as a guide for changing the type of and reducing the number of fuzzy labels, and tuning can be performed in a short time.

[0015]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing a schematic configuration of a fuzzy controller, and FIG. 2 is a block diagram of an embodiment in which FIG. 1 is more concretely embodied. The method for implementing the fuzzy controller can be implemented by either hardware (fuzzy specialized chip) or software, but here, an example of implementation by software using a general-purpose CPU will be described.

In FIG. 1, reference numeral 11 denotes a deviation calculating means for obtaining a control deviation E from the controlled variable X from the controlled object 12 and its target value S. 13 is a fuzzy arithmetic processing means,
This is because the control deviation E is received, the fuzzy operation is executed using the control rule and the membership function of each storage unit 14, and the fitness of each fuzzy variable of each input variable is newly stored in the storage unit 15. Conformity display table 15
a is provided and the function of efficiently tuning the control rule and the membership function by using the goodness of fit of the table 15a. Reference numeral 16 is a human interface device.

Next, the fuzzy controller is shown in FIG.
Will be described more specifically with reference to. In the figure, reference numeral 21 is a CP that executes a predetermined process based on program data.
It is a fuzzy arithmetic processing means using U and corresponds to the fuzzy arithmetic processing means 13 in FIG. A ROM section 23 and a RAM section are connected to the bus 22 derived from the fuzzy arithmetic processing means 21. Therefore, in this fuzzy arithmetic processing means 21, the ROM section 2 is routed via the bus 22.
3 and the RAM section are accessed, and predetermined processing is executed according to the designated program. The programs that define the operation of the controller include a system program called a scheduler program and a system standard program, and a user application program freely defined by the user.

A program (scheduler) for managing and executing user application programs and a system standard program (input / output management program) for executing input / output are mounted on the ROM section 23. On the other hand, the RAM section is roughly divided into a user application program 24A, a user work area 24B, and a system work area 24C.

This user application program 2
4A is composed of an input / output definition part and a control calculation definition part, and is defined as a part of the control calculation in the fuzzy calculation. On the other hand, the user work area 24B is an area for storing the data of the control calculation defined by the user, and stores the control rules and membership functions necessary for fuzzy control. The system work area 24C is an area where the scheduler stores information necessary for operating the system standard program and running the user application program 24A.

By the way, the tuning in the fuzzy controller is to change the form of the membership function and the contents of the rule table under the same idea as tuning the control parameter in the conventional PID controller. However, the conventional fuzzy controller is currently tuned by trial and error as described above.

Therefore, the fuzzy controller according to the present invention is provided with a fitness degree display table 24C 'in the system work area 24C and stores the fitness degree of the fuzzy variables corresponding to each fuzzy variable in this table 24C'. , For example, the arrangement shown in FIG. Here, n is the number of inputs X _i , m is the number of divisions of the membership function in one input, and μ _ij (X _i ) is the fitness of each fuzzy set.

Reference numeral 25 in the drawing sends the control deviation between the controlled variable from the controlled object 26 and the target value to the bus 22, and controls the manipulated variable after the fuzzy arithmetic processing means 21 executes the fuzzy arithmetic operation. An input / output interface leading to the target 26. Reference numeral 27 is a human interface and 28 is a communication interface, which inputs the goodness of fit of fuzzy variables corresponding to each fuzzy variable to the membership function and the goodness of fit display table 24C ', and displays the simulation result. And compatibility table 24
It has a function of outputting and displaying the goodness of fit of the fuzzy variables stored in C '.

Next, the operation of the above fuzzy controller will be described. Similar to a general process control system, the input / output interface 25 takes in the control deviation between the controlled variable from the controlled object 26 and the target value, and the fuzzy arithmetic processing means 21 outputs the manipulated variable obtained by the fuzzy arithmetic operation. Is applied to the controlled object 26. Therefore,
The input / output interface 25 controls the exchange of signals between the controller body and the controlled object 26.

On the other hand, the control rules and membership functions required for fuzzy calculation are input from the Human interface device 27 and stored in the user work area 24B via the communication interface 28.

After the necessary data is input or set as described above, the fuzzy arithmetic processing means 21 uses the control rule and the membership function for the control deviation from the input / output interface 25 according to the user application program 24A. The fuzzy operation is executed to obtain the manipulated variable, which is applied to the controlled object 26 via the input / output interface 25.

The fitness data of each fuzzy variable on the input membership function used when the fuzzy operation is performed is stored in the fitness display table 24C 'in the system work area 24C. Saved as.

Here, these table values left as information are used as a judgment material in the tuning of the fuzzy controller. Specifically, for example, an array showing the degree of conformity as shown in FIG. 3 is used, and this is arranged via the communication interface 28 to the human interface device 27.
To display. As for the display method, the time series change of the goodness of fit is treated as trend data, or the goodness of fit is integrated and treated as the frequency of use, and those data remain as numerical values,
Alternatively, a method of displaying by a graphic graph may be considered.

Therefore, by displaying the fitness data of each fuzzy variable on the input membership function stored in the fitness display table 24C 'as described above,
At the time of a simulation or a test using an actual machine, it can be used as a criterion for changing the type of membership function or reducing the number of fuzzy labels based on the conformity information of the table 24C '.

For example, when the range of input values is narrower than expected, it is possible to change the range of input values because the table value shows that the fuzzy variables at the ends of the membership function are not used at all. is there. Also,
Areas that are frequently used within the range of input values can be determined from the table values. Therefore, the controllability can be improved by increasing the number of fuzzy labels or changing the membership function type for that area. .. If many fuzzy variables that are rarely used are described in the rule table, reduce the number of fuzzy labels or change the rules so that many rules affect fuzzy inference. It can be used as a criterion for making decisions.

[0031]

As described above, according to the present invention, not only the control knowledge of the operator, which is an advantage of fuzzy control, can be expressed in a rule base, but also the type of the membership function is changed at the time of simulation or adjustment by an actual machine. It can be used as a guideline for reducing the number of fuzzy labels and, therefore, a fuzzy controller that efficiently executes control rules and tuning of membership functions can be provided.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram of a fuzzy controller according to the present invention.

FIG. 2 is a block diagram more specifically showing the fuzzy controller shown in FIG.

FIG. 3 is an array diagram showing the goodness of fit stored in the goodness-of-fit display table.

FIG. 4 is a schematic configuration diagram of a conventional fuzzy controller.

FIG. 5 is a characteristic diagram of a general membership function.

FIG. 6 is a diagram showing control rules stored in a table format in a storage unit.

[Explanation of symbols]

11 ... Deviation computing means, 12, 26 ... Control object, 13 ... Fuzzy computing device, 14, 15 ... Storage section, 21 ... Fuzzy computing processing means, ROM section 23, 24A ... User application program, 24B ... User work area, 2
4C ... System work area, 25 ... Input / output interface, 27 ... Human interface device, 28 ... Communication interface.

Claims (1)

[Claims] 1. A fuzzy controller that executes a fuzzy control operation using a control rule and a membership function and applies an operation amount obtained by the fuzzy control operation to a control target, wherein the fuzzy label is fuzzy labeled by the membership function. A goodness-of-fit display table that stores the goodness of fit of each fuzzy set of input variables corresponding to the controlled variable, and a human interface device that inputs various information for tuning the control rule and the membership function, It is provided with a tuning display means for reading out the goodness of fit of each fuzzy set stored in the goodness-of-fit display table, displaying the goodness of fit on the Fuman interface device, and tuning the control rule and the membership function. Fuzzy controller.