JPH044401A - Fuzzy control rule auto-tuning device - Google Patents

Abstract

PURPOSE:To tune a fuzzy control rule even when the difference (initial response deviation) between an ideal response and response by an initial control rule is somewhat larger by applying fuzzy inference to the inference of the consequent correcting amount of the fuzzy control rule. CONSTITUTION:A fuzzy inference means 11 for tuning operates the basic consequent correcting amount of a fuzzy control rule from a response deviation which is the difference between an ideal response value stored in an ideal response value storing means 8 and a controlled response value stored in a controlled response value storing means 7 and the varying amount of the response deviation on the basis of a membership function for tuning stored in a fuzzy knowledge storing means 10 for tuning and a fuzzy rule for tuning. The consequent section of the fuzzy control rule is corrected from the basic correcting amount obtained from the operation and the degree of materialization of the fuzzy control rule stored in a means 6 for storing the degree of materialization of the fuzzy control rule. Therefore, the fuzzy control rule can be tuned precisely, because precise correction can be performed in accordance with the response deviation and magnitude of the deviation.

Description

Detailed Description of the Invention (a) Industrial Application Field The present invention automatically adjusts and corrects fuzzy control rules so that optimal control can be achieved in a fuzzy control device that performs control based on fuzzy control rules. The present invention relates to a fuzzy control rule automatic tuning device that performs.

(b) Conventional technology Fuzzy control is based on fuzzy knowledge consisting of membership functions and fuzzy control rules, and calculates the optimal manipulated variable according to the controlled object from, for example, control (response) deviation and its difference information. Take control. As a result, nonlinear and variable gains that cannot be obtained with conventional PID (proportional, integral, differential) control can be easily realized, and highly accurate control is possible. For this reason, it has been applied to a large number of control systems.

In order to perform good fuzzy control, it is necessary to construct fuzzy knowledge suitable for the controlled object. So, for example,
[Design of Self-Adjusting Fuzzy Control Devices J (1989, Proceedings of the 5th Fuzzy System Symposium, pp. 89 to 94) describes control deviation, first-order difference of control deviation, and control deviation as antecedent variables. Take the second-order difference of the manipulated variable as the consequent variable, take the first-order difference of the manipulated variable as the consequent variable, and set the three antecedent variables as N (negative) and Z (z
In a fuzzy control device composed of fuzzy control rules obtained from the fuzzy division results into ero: zero) and P (positive: positive), after adjusting the scaling factor that normalizes the input and output values of the fuzzy control device through learning. When a control response is obtained by sampling during control operation, the fuzzy control rule is automatically tuned to correct the conclusion part (the manipulated variable in the consequent part) of the fuzzy control rule to obtain the target response. There is.

(c) Problems to be Solved by the Invention However, in the automatic tuning described above, the basic modification amount of the conclusion part of the fuzzy control rule is determined based on the ideal response waveform at the time of sampling.

The response deviation, which is the difference from the response waveform, and the change from the previous response deviation are determined only based on information such as whether they are positive, negative, or zero (the final correction amount is determined by the basic correction (the amount multiplied by the probability of each fuzzy control rule being met), it was not possible to make fine-grained corrections to the conclusion part of the fuzzy control rule according to the response deviation or the magnitude of the change in the response deviation.

Therefore, when the difference between the ideal response waveform and the control response waveform based on the fuzzy control rule before modification is large, the convergence efficiency of tuning is not necessarily good.

The present invention has been made in view of the above points, and it is possible to perform precise tuning of fuzzy control rules according to the response deviation, which is the difference between the ideal response and the control response, and the magnitude of the change in the response deviation. The present invention provides a fuzzy control rule automatic tuning device that performs automatic tuning.

(d) Means for Solving the Problems The present invention provides a fuzzy control device that controls a control response value to a preset target value based on a membership function and a fuzzy control rule. A fuzzy control rule automatic tuning device that corrects the following: control response value storage means for storing a control response value; rule establishment degree storage means for storing the degree of establishment of the fuzzy control rule with respect to the obtained control response value; ideal response value storage means for storing an ideal response value in control of the fuzzy control device; A tuning fuzzy knowledge storage means for storing a tuning membership function and a tuning fuzzy rule for modifying the value of the consequent of a fuzzy control rule, and a tuning membership stored in the tuning fuzzy knowledge storage means. the degree of fulfillment of the fuzzy control rule stored in the rule degree storage means based on the function and the fuzzy rule for tuning;
The fuzzy control rule is calculated based on the response deviation which is the difference between the ideal response value stored in the ideal response value storage means and the control response value stored in the control response storage means and the magnitude of the change in the response deviation. and a tuning fuzzy inference means for modifying the value of the consequent.

(E) The fuzzy inference means for action tuning uses the ideal response value stored in the ideal response value storage means based on the membership function for tuning and the fuzzy rule for tuning stored in the fuzzy knowledge storage means for tuning. The basic amount of correction for the consequent of the fuzzy control rule is calculated from the response deviation, which is the difference between the control response values stored in the control response storage means, and the magnitude of the change in the response deviation. The consequent part of the fuzzy control rule is corrected based on the obtained basic modification amount and the degree of fulfillment of the fuzzy control rule stored in the rule degree storage means. As a result, precise correction is made according to the response deviation and the magnitude of the response deviation.

(F) Embodiment FIG. 1 is a schematic diagram of an embodiment of the fuzzy control rule automatic tuning device of the present invention and a fuzzy control device to be tuned.

(1) is a fuzzy control device, which uses a fuzzy control knowledge base (2) in which fuzzy control rules and membership functions are stored; A fuzzy inference unit (
3).

In the fuzzy control device (1), the fuzzy inference unit (3) performs fuzzy control based on the control deviation el between the set value r and the control response y at the current sample time j, and the first-order difference Δe of the control deviation. The first-order difference Δ of the manipulated variable is calculated based on the fuzzy control rule stored in the knowledge base (2).
U, is calculated, and further, the manipulated variable U is determined by uj=u) and Δuj. Then, by adding this operation amount U to the controlled object (4), the controlled object (4) is controlled.

(5) is a fuzzy control rule automatic tuning device that automatically tunes the fuzzy control rules stored in the fuzzy control knowledge base (2) of the fuzzy control device (1). (
The rule establishment degree storage table (6) serves as a rule establishment degree storage means for storing the degree of establishment of the fuzzy control rule that was established for the control responses from sample time j of the current sample up to S samples ago, and similarly the latest sample time j to S
Response waveform storage table (7) (12th
(see figure), an ideal response waveform storage table (8) (first
(See Fig. 3, Fig. 13 also shows the response deviation e, 6), which is stored in the response waveform storage table (7) from the ideal response value Yi° stored in the ideal response waveform storage table (8). The control performance evaluation unit (9) calculates a response deviation e,6 obtained by subtracting the actually obtained control response value, and a difference Δe1° between this response deviation e1° and e,−1° before S samples. ing. Furthermore, this fuzzy control rule automatic tuning device (5) has a control performance evaluation section (9).
) Tuning fuzzy rules and members for modifying the value of the consequent part of the fuzzy control rule based on the response deviation e, 0 obtained by A fuzzy knowledge base for tuning (10) as a fuzzy knowledge means for tuning in which ship functions are stored, and a fuzzy control rule is created based on the fuzzy rules for tuning and membership functions stored in the fuzzy knowledge base for tuning (10). Tuning fuzzy inference unit (1) as a tuning fuzzy inference means for modifying the consequent part
1).

First, the operation of the fuzzy control device (1) will be explained. In this embodiment, for simplicity, it is assumed that a simple inference in which the consequent part is a real value is used in the fuzzy control rule.

Now, the fuzzy control rule sets the control deviation el to the antecedent variable.
Let us take the first-order difference 80 of the control deviation and the first-order difference ΔU of the manipulated variable as the consequent variable.

Although the fuzzy division of the control deviation e and the first-order difference Δe1 between the control deviations is arbitrary, here we divide the control deviation el and the first-order difference Δe between the control deviations into seven parts, as shown in Figure 2. (
NB: negative big, NM: n
aggressive mediumSN S: nega
tive small, Z 0: zero, P
S: positive small, PM
: positive medium, PB:
positive, big), and set 13 fuzzy control rules as shown in FIG. It's okay to add more rules to the blank spaces in Figure 3.

The fuzzy control rule shown in FIG. 3 is expressed as follows.

R1: IF eIs NB, △ei is ZO,
Tt(EN△+g is NBR2:IF ei is
NM+ΔeIis ZO, THENΔJ is NM
R3: IF ei is NS, Δel is ZO
, THENΔul is N5R4:rFel is
ZO, Δe) is NB, THEN △ui is
NBRll:IF e) is PS, △ei i
s ZO, THENΔJ is PSR12:IF e
i is PM, △eIis ZO, Tt(ENΔu
Iis PMR13: IF el is PB, Δei
is ZO, THENΔJ is PB However, the latest (
Let J be the sample point in time (current), and e+=r−yi +
Δe i=e j −e i−.

ΔuI=u+uj−1 r: set value, y, two-control response, e, two-control deviation, Δe,
The first-order difference between the two control deviations, Δu, is the first-order difference between the manipulated variable U.

The membership function of the antecedent part is as shown in FIG. 2, and the membership function of the consequent part is shown in FIG. In this example, the membership function of the consequent part is a real-valued number (i=
1 to 7).

When the inputs ei, Δe] of the fuzzy control device are given, the output Δu1 is obtained by the following equation.

However, w is the degree of rule establishment of the first rule with respect to ei and Δel.

From the first-order difference Δul of the manipulated variables obtained in this way and the manipulated variables u, -1 at the previous sample time point J-1,
Add these to find the manipulated variable U. Then, this manipulated variable u1 is added to the controlled object (4) to control the controlled object (4).

Next, tuning of fuzzy control rules will be explained. Tuning Fuzzy Control Rules 11 - where is the observed control response y1 at the current sample time j
is the ideal response yI that is set in advance and stored in the ideal response waveform storage table (8). In order to match the response deviation e, 6, and S
The change in response deviation between samples Δe is determined by fuzzy inference according to the sign of 6 and the magnitude of its value, and is performed by increasing or decreasing the real value of the consequent part of the control rule.

Here, eI. = '/ + Y r'Δe, 0
=e, Oe, -, 6eI6: Response deviation, Δej
': Change in response deviation between S samples, y, °: Ideal response, yI: Control response.

For example, taking FIG. 8A as an example, in FIG. 8A, "At the current moment j, the response deviation e, ° is NB, and the response deviation has further decreased compared to before S samples (Δe, 6 (N) tends to move away from the ideal response toward a smaller value, so it is thought that the amount of manipulation before the S sample was too small, and the value of the consequent of the control rule used before the S sample was increased. Indicates cases in which the amount must be increased.

Expressing this using fuzzy rules for tuning, at time j, IF e , a is NB, Δe jois N,
THENΔh+1sPB However, Δ11. : Expressed as the amount of modification of the value of the consequent of the i-th fuzzy control rule.

The fuzzy division of the variables in the antecedent and consequent parts of the fuzzy rule for tuning is still arbitrary, but the response deviation e
, Q is shown in FIG. 5, and FIG. 6 is a fuzzy partition example of the first-order difference Δe of the response deviation, 6. Furthermore, as shown in FIG. 7, the membership function of Δh in the consequent part is replaced with a real value p, since simplified inference is also used in fuzzy inference for tuning.

That is, since PB corresponds to R7, the above tuning fuzzy rule is IFe. is NB, Δe,. is N, THE
NΔhlisp+.

The method of generating the control response yl for the ideal response y+' corresponding to all the fuzzy divisions shown in FIGS. 5 and 6 is explained in Section 8.
Shown in Figures A to O. The tuning fuzzy rules corresponding to FIG. 8 are expressed in a table format as shown in FIG. 9.

Incidentally, the fuzzy rules for tuning need not be described for all cases as shown in FIG. 9, but may be described only for typical patterns as shown in FIG. 10, for example.

The tuning fuzzy inference unit (11) uses the membership functions (FIGS. 5 to 6) and FIG. 9 (or
The response deviation e calculated in the control performance evaluation unit (9) based on the fuzzy rules for tuning shown in Figure 0),
6, and the change in response deviation (first-order difference in response deviation) Δ
When e, 6 is given, by simple inference, the modification amount Δh1 of the consequent part of the control rule is given as fuzzy rules for tuning e,
+1, Δe, and the degree of rule establishment for o is μ).

Then, the value h1 of the consequent part of the fuzzy control rule is modified by the following equation at the current sample time j for all fuzzy control rules used to calculate the manipulated variable before S samples.

hlNEW=hlOLD×W(I-1)·Δh+where, wL+-“”: The degree of fulfillment hl of the first fuzzy control rule at the time of (j-s). 1'D = value of the consequent part of the fuzzy control rule used at time point (j-8), NEW, value of the consequent part of the fuzzy control rule modified at time point 5.

w”−“) is calculated by the fuzzy inference unit (
3) is determined for each fuzzy control rule, and this is stored in the rule establishment degree storage table (6). FIG. 11 shows an example of the contents of the rule establishment degree storage table (6).

In this manner, the value of the consequent of the fuzzy control rule used at time (j-s) in the fuzzy control knowledge base (2) is modified.

Tuning of such fuzzy control rules is carried out, for example, by tuning the controlled object (4
), real-time tuning is repeatedly performed using the fuzzy control rule after the previous tuning as the initial rule until the criterion is satisfied. Note that δ is the value of the termination criterion.

(G) Effects of the Invention As is clear from the above description, the present invention applies fuzzy inference to the inference of the amount of modification of the consequent of the fuzzy control rule, thereby determining the response deviation e,° and the change in the response deviation. The fuzzy control rule is tuned more precisely according to the magnitude of the minute Δe,° (first-order difference in response deviation).

Then, the difference between the ideal response and the response according to the initial control rule (
This makes it possible to perform tuning even when the initial response deviation (initial response deviation) is large to some extent, thereby improving the robustness and convergence efficiency of tuning.

[Brief explanation of the drawing]

FIG. 1 is a schematic configuration diagram of an embodiment of the fuzzy control rule automatic tuning device of the present invention and the fuzzy control device to be tuned, and FIG. 2 is a diagram of members of the antecedent part of the fuzzy control rule according to the embodiment of the present invention. A diagram showing the ship function,
FIG. 3 is a diagram showing a fuzzy control rule according to an embodiment of the present invention, FIG. 4 is a diagram showing a membership function (real value) of the consequent part of the fuzzy control rule according to an embodiment of the present invention, and FIG. 6 and 6 are diagrams showing membership functions of the antecedent part of the fuzzy rule for tuning according to an embodiment of the present invention,
FIG. 7 is a diagram showing the membership function (real value) of the consequent part of the fuzzy rule for tuning according to an embodiment of the present invention, and FIG. 8 is a diagram showing the state of the control response with respect to the ideal response.
9 and 10 are diagrams showing fuzzy rules for tuning according to an embodiment of the present invention, FIG. 11 is a diagram illustrating an example of a rule establishment degree storage table according to an embodiment of the present invention, and FIG. A diagram showing an example of a response waveform storage table according to an embodiment of the invention. FIG. 13 is a diagram showing an example of an ideal response waveform storage table according to an embodiment of the invention. (1) Fuzzy control device, (2) Fuzzy control knowledge base, (3) Fuzzy inference unit, (4)
...Controlled object, (5)...Fuzzy control rule automatic tuning device, (6)...Rule satisfaction degree storage table (rule fulfillment degree storage means), (7)...Response waveform storage table (control response value storage means), (8)...Ideal response waveform storage table (ideal response value storage means), (9)...
- Control performance evaluation unit, (10)...Tuning fuzzy knowledge base (tuning fuzzy knowledge storage means), (11)...Tuning fuzzy inference unit (tuning fuzzy inference unit).

Claims (1)

[Claims] (1) In a fuzzy control rule automatic tuning device that corrects a fuzzy control rule for a fuzzy control device that controls a control response value to a preset target value based on a membership function and a fuzzy control rule. , a control response value storage means for storing a control response value, a rule fulfillment degree storage means for storing the degree of fulfillment of the fuzzy control rule with respect to the obtained control response value, and an ideal response value for controlling the fuzzy control device. and correcting the value of the consequent part of the fuzzy control rule according to the response deviation that is the difference between the ideal response value and the control response value and the magnitude of the change in the response deviation. and a tuning fuzzy knowledge storage means for storing a tuning membership function and a tuning fuzzy rule; and the establishment of the rule based on the tuning membership function and the tuning fuzzy rule stored in the tuning fuzzy knowledge storage means. response deviation which is the difference between the ideal response value stored in the ideal response value storage means and the control response value stored in the control response storage means; and fuzzy inference means for tuning that corrects the value of the consequent of the fuzzy control rule based on the magnitude of the change in response deviation.