JPH0378003A - Fuzzy controller - Google Patents

Abstract

PURPOSE:To perform proper feedback control in both a transient state and a stationary state by providing a means which varies the scale of a membership function according to an input variable value level. CONSTITUTION:When there is the input value signal of a controlled variable Y form a controlled system 5, the deviation (e) of the input value signal from a desired value (r) is calculated and differentiated. Then the mean deviation E of past (n) values is calculated to calculate the scale S(e) of the deviation (e), the scale S( e) of the differentiated deviation e, and the scale S( u) of manipulated variable variation u respectively. The scale variation part 10 varies the scales of respective membership functions to the calculated scales. Thus, a fuzzy reasoning part 3 performs fuzzy reasoning by using the deviation (e) and differentiated deviation e as input variable values after the scales of the respective member ship functions are varied.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a fuzzy controller used for controlling a feedback system.

(Prior art) In a fuzzy controller, the degree of fitness for each antecedent membership function of a given input value is calculated, and the antecedent degree of fitness, which is the minimum value of these degrees of fitness, is calculated for each rule. Truncate the consequent membership function in each rule based on the antecedent suitability, then superimpose the truncated consequent membership functions, and enter the final value corresponding to the center of gravity. It is outputting for the value.

Fig. 3 shows the functional block configuration when the fuzzy controller is used to control a feedback system, where l is a deviation generation section, 2 is a differentiation circuit section, 3 is a fuzzy inference section, 4 is a previous operation amount storage section, and 5 is the controlled object. Deviation generating part l
Then, the deviation e between the controlled variable Y and the target value r is calculated, and the differentiator circuit section 2 calculates the deviation differential Δe, which is the variation amount of the deviation e over time, and the fuzzy inference section 3 converts the input into the deviation e.
, fuzzy inference is performed as the differential deviation Δe to output the manipulated variable change ΔU, and the manipulated variable U obtained by adding the previous operation 1u given from the previous manipulated variable storage unit 4 to the manipulated variable change ΔU is given to the controlled object 5. and the resulting control IY
is output from the controlled object 5.

FIG. 4 is a diagram showing control waveforms in feedback control, and shows that the controlled amount Y approaches the target value r as time passes, and the deviation e is large in the control process and the controlled amount Y and the target value r. A state that is not consistent with the value r is defined as a transient state, and a state that is balanced is defined as a steady state.

(Problems to be Solved by the Invention) As described above, when performing feedback control using fuzzy inference, there are the following problems.

Figure 5 shows the fuzzy rule matrix used for temperature control reasoning, and Figure 6 shows the membership function regarding the deviation e used for temperature control, where the deviation e is r-100°C.
There are seven label membership functions in the range from `` to 1+too°C''.

In a transient state, the range of change in the value of the deviation e given sequentially is large, so the change in the degree of fitness of the antecedent part is also remarkable, but in a steady state, the value of the deviation obtained sequentially is [0
°CJr + 1°CJr - 2°C" and its change width is small, so there is almost no change in the antecedent fitness, and correspondingly, the output value also hardly changes,
In other words, a steady-state deviation remains and normal control in a steady state cannot be performed.

In order to solve the above problem, it is possible to use a membership function with a small width so that the antecedent fitness changes even when the variation in the value of deviation e is small. , in which case the number of membership functions,
The number of rules corresponding to this became enormous, and setting them required time and effort, and the inference time also increased, making it not a desirable method.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a fuzzy controller that enables appropriate control in each of the transient state and steady state in feedback control.

(Means for Solving the Problem) In order to achieve such an object, the present invention uses a fuzzy controller to calculate output variable values for given input variable values by fuzzy inference based on fuzzy rules and membership functions. The present invention is configured to include fuzzy inference means for outputting an output, and membership function scale changing means for changing the scale of the membership function based on the input variable value level.

(Operation) According to the above configuration, fuzzy inference is performed based on the membership function whose scale is changed in accordance with the input value level.

(Example) Hereinafter, an example of the present invention will be described in detail with reference to the drawings. FIG. 1 is a functional block diagram showing the overall configuration of the fuzzy controller of the present invention, and the same parts as those shown in FIG. 3 of the conventional example are given the same numbers.

The present invention includes a scale changing unit 10 as a membership function scale changing means that detects a deviation e level that is an input variable value level and changes the scale of a membership function based on the detected value.

The fuzzy controller shown in this embodiment is of a type in which the inference process is processed by software, and the fuzzy inference operation in the fuzzy inference unit 3 as a fuzzy inference means,
The scale changing operation in the scale changing section 10, the differential processing operation in the differentiating circuit section 2, the deviation calculating operation in the deviation generating section 1, etc. are performed by the operation of a CPU constituted by a microcomputer. Fuzzy rules, membership functions, etc. used in inference are stored in memory, and the previous operation storage section 4 is also constituted by memory.

Next, its operation will be explained with reference to the flowchart in FIG.

When there is an input value signal of the controlled variable Y from the controlled object 5 (step Sl), the deviation e between the input value signal and the target value r is calculated (step S2), and the deviation e is differentiated (step S1). S3). Next, calculate the average deviation E over the past n times (step S4), as shown in the following formulas:
For each constant, 2. By multiplying by
U) are calculated (step S5).

5(e) = ・E S (Δe) = ・E S (Δu ) = K 3 ・E Each of the above coefficients, , , KG are determined by the controlled object 5, and when the scale is fixed If the optimal scale for transient response or steady state is found, it is possible to calculate from that.

Next, the scale of each membership function is changed to the calculated scale (step S6). The reason why the average deviation E is used in the above is to remove noise.

By changing the scale of the membership function in this way, for example, in FIG.
The scale of the membership function in is r-100°
In the case of a large case of 200°C ranging from "C" to "1+100°C", the deviation e of "0°C" [+1°CJr-26C] is all located approximately in the center of the membership function ZH, but the scale is "- 5°C” to “+5°C”
If it is reduced to 00C, "0° CJr+l.

The deviation e of ``C, H-2°C'' will be distributed in each membership function, and as a result, the degree of fitness of the antecedent part during fuzzy inference will also change significantly.

In the above, if the deviation e obtained this time is larger than the average deviation E, the current deviation e is set to E. This improves responsiveness when the deviation e suddenly increases due to disturbances, target value changes, etc. , and if the average deviation E is smaller than the set minimum average deviation Em + n, then the average deviation E
is Emin. This is to prevent hunting since the scale also becomes zero when the average deviation E becomes zero after settling.

After the scale of each membership function is changed as described above, fuzzy inference is performed using the deviation e and the differential deviation Δe as input variables, and the manipulated variable change ΔU is output (step S7). The manipulated variable change ΔU is added to the previous manipulated variable U (step S8), and after the current manipulated variable is stored (step S9), the manipulated variable U is output (step 510).

The average deviation E in the above is large during a transient response and small during a steady response.On the other hand, the combination of the scales of the antecedent membership function regarding the input and the consequent membership function regarding the output is as follows. There are four possible options: A, B, C, and D.

A: Large input scale, large output scale B: Large human scale, small output scale C: Small input scale, large output scale, small scale for two people, small output scale And in the case of A above, the output scale is large, so ,
The amount of change in the manipulated variable ΔU becomes larger, the change in the manipulated variable U becomes larger, and the response during transient response is faster.

On the other hand, during a steady response, even if there is a small deviation e, the input is regarded as ZR, and since there is almost no output of the manipulated variable change ΔU, a steady deviation remains.

In case B, since the output scale is small, the amount of change in the manipulated variable ΔU is small, the change in the manipulated variable U is small, and the responsiveness during transient response is slow. On the other hand, during steady response, a steady deviation remains as in case A.

In case C, as in case B, the responsiveness during transient response is slow, whereas during steady response hunting occurs because the manipulated variable change ΔU is output even if the deviation e is small.

In case D, the response during transient response is fast as in case A, but on the other hand, during steady response, a small manipulated variable change ΔU is output when the deviation e is small, so hunting is unlikely to occur. do not have.

From the above, it can be determined that A or C scale setting is preferable for transient response, and D is preferable for steady response.
The scale may be set according to such content.

(Effect of the Invention) Therefore, according to the present invention, fuzzy inference is performed based on the membership whose scale is changed corresponding to the input value level. For example, in feedback control or the like, it is now possible to perform appropriate control in both a transient state and a steady state where there is a difference in input value level, without causing any problems.

[Brief explanation of drawings]

1 and 2 relate to an embodiment of the present invention, in which FIG. 1 is a functional block diagram of a fuzzy controller, and FIG. 2 is a flowchart for explaining the operation. FIG. 3 is a functional block configuration diagram of a conventional example of a fuzzy controller, FIG. 4 is an explanatory diagram showing general control waveforms in feedback control, and FIG. 5 is an explanatory diagram of a general fuzzy rule matrix used in feedback control.
FIG. 6 is an explanatory diagram showing membership functions related to deviation. 3... Fuzzy inference section (fuzzy inference means), 10.
- Scale change unit (Membership function scale change means). Figure Figure

Claims (1)

[Claims] (1) Fuzzy inference means that outputs an output variable value for a given input variable value by fuzzy inference based on a fuzzy rule and a membership function, and changing the scale of the membership function based on the input variable value level. A fuzzy controller comprising: membership function scale changing means;