JPH0335301A - Fuzzy controller - Google Patents

Abstract

PURPOSE:To compensate a part where the satisfactory accuracy is not secured by a fuzzy inference device with use of a compensator by using an adder which adds the output of the fuzzy inference device to the output of the compensator containing a linear or nonlinear control part in the prescribed ratio. CONSTITUTION:A fuzzy controller consists of a fuzzy inference device 1, a PID compensator 2, and an adder 3. The compensator 2 contains a linear PID control part, and the common inputs theta1 - thetan are inputted to both the device 1 and the compensator 2. The adder 3 decides the goodness-of-fit values of outputs of the device 1 and the compensator 2 and adds both values together. Thus the compensator 2 applies the proper compensation to the output of the device for a subject to which the highly accurate control is difficult with the fuzzy inference only. As a result, the output accuracy of a fuzzy controller can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Industrial Application Field The present invention relates to a fuzzy control device in which a fuzzy inference device is combined with a compensator consisting of a linear or nonlinear control section.

(b) Prior Art In a conventional fuzzy control device, the control section was composed of only one fuzzy inference device, and the output of the fuzzy inference device was directly used as the operation amount for the output section such as an actuator.

Such a fuzzy control device is generally convenient when the model of the object to be controlled is not well understood, but when the model of the object is well known, a control system suitable for it,
For example, a PID control device is more preferable.

(C) Problems to be Solved by the Invention However, when the conventional fuzzy control device, PID control device, etc. described above are used individually, the following problems arise.

■When the control unit is composed of only fuzzy inference machines (1) Since fuzzy operations are structurally nonlinear, fuzzy operations must be used even when an accurate linear model near the stable point is known. Therefore, accuracy cannot be achieved.

(2) If fuzzy operations were to achieve performance equivalent to or better than a linear controller, it would be necessary to manipulate membership functions and fuzzy rule maps, which would require a great deal of effort until a decision is made.

■When the control unit consists of only a linear control unit such as a PIDIJ control unit.

(1) Near the stable point, there are linear models and nonlinear models, and even if control is successful, at points away from the stable point, the nonlinearity becomes stronger and it often becomes impossible to control well with either model.

(2) For example, if the nonlinearity of the controlled object is known in advance, it is possible to use calculations or maps to compensate for the nonlinearity, but while this method allows for good control, it takes a long time to calculate. , has the disadvantage of increasing the load on the CPU.

As mentioned above, even when controlling the controlled object only by the fuzzy inference machine, PID! Even when controlling only with other linear controllers such as II controllers, there is a problem in that highly accurate and efficient control is not possible in all conditions.

An object of the present invention is to combine a fuzzy inference device with a compensator consisting of a linear or non-linear control section, and to compensate for the parts where the accuracy cannot be achieved with IT and II control in the fuzzy inference device, thereby solving the above-mentioned drawbacks. The purpose of this invention is to provide a fuzzy control device that solves the problem.

(d) Means for Solving the Problems This invention includes a fuzzy inference device, a compensator including a linear or nonlinear control section, and an output of the fuzzy inference device and an output of the compensator that are added at a predetermined ratio. It is characterized by comprising an adder.

Further, a fuzzy inference device, a compensator including a linear or non-linear control section, and a fitness value calculator that calculates a fitness value of the output of the fuzzy inference machine and a fitness value of the output of the compensator from fuzzy operation contents; It is characterized by comprising an adder that adds the two matching values.

(e) Effect In the invention according to claim 1, in order to configure a control device by combining a fuzzy inference device and a compensator consisting of a linear or nonlinear control section, the ratio of their outputs is appropriately set. , for control objects that have strong nonlinearity or for which precision cannot be achieved with PID control, etc.
The fuzzy control by the fuzzy inference machine can be made to work strongly, and the control by the compensator can be made to work strongly for a strongly linear controlled object whose model is well known. Further, even for a controlled object having a mixture of linear and nonlinear characteristics, accuracy can be improved by appropriately selecting the settings of the ratios.

In addition, by providing a fitness value calculator that calculates the fitness value of the output of the fuzzy inference device and the output of the compensator from the contents of the fuzzy calculation, compared to the case where the ratio of adding the outputs of both is fixed, It becomes possible to perform highly accurate control on any control object, and even on control objects whose linear and nonlinear characteristics change.

(f) Embodiment FIG. 1 shows a professional fuzzy control device according to an embodiment of the present invention.
7 diagrams are shown. In this embodiment, the fuzzy control device includes a fuzzy inference device 1, a PAD compensator 2, and an adder 3. In this embodiment, the compensator 2 is a linear PI
It consists of a D control section, a fuzzy inference device 1 and a PID compensator 2.
The common human power θ, ~θ7 is entered in . The adder 3 determines the fitness value of the output of the fuzzy inference device l and the output of the compensator 2 (the fitness degree of the fuzzy inference device l and the fitness degree of the compensator 2) according to a preset ratio, and these fitness values are will be added.

In this embodiment, in the adder 3, the fuzzy inferencer l
Since the fitness of the PrD compensator 2 and the fitness of the PrD compensator 2 are set in advance as a constant ratio, the fuzzy inferencer 1. The output of the PID compensator 2 is multiplied to obtain respective adaptive values, and these adaptive values are added and output. Therefore, when the target is difficult to control with high precision using fuzzy inference alone, the accuracy of the output of the control device can be increased by adding appropriate compensation to the output of the fuzzy inference device using the PID compensator 2. . Also, on the contrary, P
For objects that can hardly be controlled by the ID compensator 2 alone, by appropriately adding the control amount based on fuzzy inference to the control amount of the PID compensator 2, highly accurate control becomes possible in the same way as described above. In addition, if necessary, PID
It is also possible to set the amount of compensation from the compensator 2 to O using a switch or the like, so by turning on the P■DIili compensator 2 only when the fuzzy inferencer does not suit the target model, higher performance can be achieved. Precise control becomes possible.

FIG. 2 shows a further preferred embodiment of the invention.

In terms of configuration, the difference from the embodiment shown in FIG. This is the point where the device suitability value calculator 5 is placed. These fitness value calculators 4 and 5 are respectively fuzzy inference machines 1. PID compensator 2
Calculate the fitness value of the output (fitness of fuzzy inference device 1.PID compensator 2). The information for calculating the fitness value is obtained from the fuzzy reasoner l. That is, each fitness value calculator 4
, 5 calculate respective fitness values from the contents of the fuzzy operation. With such a configuration, it is possible to precisely set the amount of compensation for variations in the model to be controlled.

FIGS. 3(A) and 3(B) are diagrams for explaining the operation of this embodiment.

The horizontal axis of FIG. 2A shows the deviation e, which is human power information, and the vertical axis shows the degree of adaptation of the fuzzy inference device 1 and the PID compensator 2. In addition, the horizontal axis of the same figure (B) indicates the deviation e,
The vertical axis shows the output of adder 3.

As shown in FIG. 3(A), in the range of deviation e=±5v or more, the fitness of the fuzzy inference device 1 becomes 1, and the fitness of the PID compensator 2 becomes 0. Therefore, in this range, only the fuzzy inference unit l is controlled. And -5>
The compensation of the PID compensator 2 starts to be effective in the region of e>+5.

Then, when e=0, the fitness of the PID compensator 2 becomes 100%.

Let us look at the change in the output of the adder 3 due to such a change in the degree of adaptation in FIG. 3(B). The hatched area in the figure indicates the range to which the PID compensator 2 contributes. Thin line F
indicates the output of fuzzy inference unit I, and thick vAO indicates the output of adder 3.
This is the output of As shown in the figure, the sensitivity in the range S is very small due to the fuzzy output F alone, but due to the compensation (contribution) of the PID compensator 2 shown by the hatching P, the sensitivity in the range S, that is, in the vicinity of e=o Sensitivity increases and accuracy improves. When e=o, both the fuzzy output F and the P10 compensation amount are 0, so even if the fitness of the PID compensator 2 is 1, the output O is 0. On the other hand, −5>e
>What a P in the range of +5! Even if the compensation amount of the D compensator 2 is large, its fitness is 0, so the output O matches the fuzzy output F.

As shown in FIG. 3(A), the fuzzy operation contents of the fuzzy inference device 1 are used to change the fitness degree of the fuzzy inference device 1 and the fitness degree of the PID compensator 2 according to the magnitude of the deviation e. be done. In this method, as will be described later, it is possible to use the degree of fitness (of the consequent part) of a particular fuzzy rule obtained by m1ni calculation of the antecedent part of the fuzzy rule. For example, if the variables of the antecedent part of the fuzzy rule are e and its change e', then to obtain the change in fitness as shown in Figure 3 (A), If e=o, e' = Othen F=ZR−・−−
---------- Use the m1ni calculation result of the antecedent part when the variables e and e' are applied to the rule (■). That is, the complement of 1 of the m1nt operation result is given to the fuzzy adaptation value calculator 4, and the m1ni operation result is given as it is to the compensator adaptation value calculator 5. Here, the reason why the complement of 1 of the fitness value of the consequent of the fuzzy rule of the above equation 0 is given to the fuzzy fitness value calculator 4 is because the sensitivity of the fuzzy output becomes extremely low near e=0. This is because the PID compensator 2 compensates for this. In other words, in the vicinity of e=0, it is necessary to lower the fitness of the fuzzy inferencer l and increase the fitness of the PID compensator 2, so the adaptation (of the consequent part) of the fuzzy rule of equation 0 above This is because when the degree is high, it is necessary to conversely lower the degree of fitness of the fuzzy inference device 1 and increase the degree of fitness of the PID compensator 2.

As mentioned above, by constantly determining the degree of compatibility between the fuzzy inference device 1 and the PID compensator 2 depending on the content of the fuzzy calculation, the output ○ of the adder 3 is calculated as e=
In the vicinity of 0, the compensation of the PID compensator 2 is greatly influenced.

According to such control, the compensation amount of the PID compensator 2, which is a linear control section, becomes large near the stable point of e=o, so that sufficient sensitivity cannot be obtained with the fuzzy inference device 1 alone. Sensitivity increases near e=0 (sufficient accuracy cannot be obtained) and accuracy improves.

Next, FIG. 4, which describes the case where the above fuzzy control device is actually applied to a motor position control device, shows the configuration of the motor position control device. The motor 10 has a position detection device 1). A speed detection device 12 is connected, and a deviation calculator 13 obtains the outputs θ and ■ of these detection devices, and generates a target value θ1 . ■Calculate the deviation from. Deviation e between θ and θ and deviation e′ between V and V
is given to the fuzzy control device 15 as fuzzy human power information, its output is gain amplified by the amplifier 16, and outputted to the motor IO as a manipulated variable.

Further, the fuzzy control device 15 is connected to a gain setter 17 of the PAD compensator 2 and a fuzzy rule and membership function setter 18.

In the above configuration, fuzzy rules and membership functions are set for the fuzzy inference device 1 of the fuzzy control device 15 by the fuzzy rule and membership setting device 18. The membership functions are a membership function corresponding to the antecedent part and a membership function corresponding to the consequent part of each rule. Figure 5 shows the fuzzy rules expressed as a matrix, Figure 6 (A) shows the membership function of the antecedent part of each rule, and Figure 6 (B) shows the membership function of the consequent part of each rule. For the compensator 2, a gain for each control device of the PID is set by a compensator gain setter 17.

FIG. 7 is a signal waveform diagram of the motor position control device having the above configuration. It is assumed that the target rotation angle θ4 and the target rotation speed ■ draw a speed trapezoidal curve as shown in (A) and (B) of the figure. Target value θ1. ■.

In contrast, as shown in FIGS. 7(C) and (D), the current value θ,
When ■ changes, the deviation is shown in the same figure (E).

The result is as shown in (F). This deviation angle e and deviation rotational speed e' become input to the fuzzy control device t15.

Now, assuming that the original point is point a in FIGS. 7(E) and 7(F), the calculations in the fuzzy inference unit 1 are as shown in FIG. 8. The membership function composite waveform of the consequent part is as shown in FIG.

Here, the purpose of the fuzzy reasoner 1 is e=0゜e'=0
Therefore, the fitness value of the output of the fuzzy Hfi logic unit 1 is obtained from the fitness of Rule 5 (or the consequent of Rule 5). Therefore, with the calculation contents shown in FIG. 8, the fitness of the fuzzy inference device 1 is 0.5.

On the other hand, if the configuration of the PID compensator 2 is as shown in FIG.
(0,4), (0,1) (however, the deviation Jed
t is -25), the output of the adder in the PID compensator 2, that is, the output of the PID compensator 2, 0UT2, is 0.5. +1.0+2.5=4.0.

Since the fitness of the fuzzy inference device 1 is 0.5 from the calculation of rule 5 in FIG. 8, the gain fuzzy inference output (determined value in FIG. 9) OUTI of the fuzzy inference device 1 is 1.
5, the fuzzy fitness value calculator 4 performs the following calculation.

1.5X (1-0,5)=0.75... ■In the above equation (1), (1-0,5) is the fitness of the fuzzy ifI logic 0.
It is the complement of 5 to 1.

Further, the compensator adaptation value calculator 5 calculates the following equation.

4, OXo, 5=2.0... ・■ Adder 3 adds the results of the above formulas ■ and ■ to 0.
75+2.0=2.75), and the value 2°75 is outputted to the amplifier 16 as a fuzzy output OUT compensated by the PID compensator 2.

By the above operation, the PID compensator 2 can compensate for the extremely low sensitivity of the fuzzy inference output near the target point of e=o. On the other hand, for example, e
=2.5. When e' = 7.5, fuzzy calculations are performed using rules LL and L4, and the composite waveform of their consequent is as shown in Figure 1). Therefore, in this case, the degree of suitability of rule L5 (the degree of suitability of fuzzy inference device l) is O. First of all, the fuzzy fitness value calculator 4 performs a calculation in which the output 0UTI of the fuzzy inference unit 1 is multiplied by (1-0). Also, the output of the PID compensator 2 is 0.2X2.5+0.4X7. 5+2. 5=6. 0, but since the fitness of this compensator 2 is O, this P
The ID compensator has no effect at all. Therefore, the output OUT of the adder 3 at this time is the defuzzify output (rJT1 planting) -4,0 in Figure 1).As in the above operation,
By combining the fuzzy inference machine and the PID compensator 2, which is a linear control unit, and by using the fuzzy fitness value calculator 4 and the compensator fitness value calculator 5, the magnitude of the human variables of the fuzzy inference machine 1 can be adjusted. The degree of compensation of the PAD compensator can be set to an appropriate level according to the following. That is, in the above embodiment, as shown in FIG. 3(B), e
A fuzzy output with low sensitivity (low accuracy) near the stable point of =0 can be made high sensitivity (high accuracy) by compensation by the PID compensator 2. Furthermore, for a nonlinear model located far from a stable point, the degree of control of the fuzzy inference device, which can respond to the nonlinear model with high precision, can be strengthened.

In the above embodiment, a PID compensator, which is a linear control unit, was used to compensate the fuzzy output, but depending on the object to be controlled, a nonlinear control unit or other fuzzy inference device may be used instead of the PID compensator. You can also use

(Effects of the g1 invention As described above, this invention makes it possible to achieve accurate control, which was difficult to achieve with only one fuzzy inference device, and also to realize models that were impossible with conventional linear controllers alone. This makes it possible to control areas to be controlled that cannot be controlled using the Adaptive Value Calculator.In other words, it has the advantage of being able to construct a control system with a very wide range of applicability.Also, by using the adaptive value calculator, even more accurate control is possible. Moreover, the scope of application becomes wider.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a fuzzy control device according to a first embodiment of this invention, FIG. 2 is a block diagram of a fuzzy control device according to a second embodiment of this invention, and FIGS. 3 (A) and (B) is the second
4 is a block diagram of a motor position control device to which the fuzzy control device of the above embodiment is applied, FIG. 5 is a diagram showing fuzzy rules, and FIG.
A) (B) is a diagram showing the membership function of the antecedent part and the membership function of the consequent part of the fuzzy rule, Fig. 7 (A)
~(F) is a signal waveform diagram of the motor position control device, No. 8
The figure shows an example of fuzzy calculation contents, Figure 9 shows the composite waveform of the consequent during fuzzy calculation, and Figure 1O shows the PI
The configuration diagram of the D compensator, Figure 1) is a diagram showing a composite waveform of the consequent part during fuzzy calculation.

Claims (2)

[Claims] (1) A fuzzy control device comprising: a fuzzy inference device; a compensator including a linear or nonlinear control section; and an adder that adds the output of the fuzzy inference device and the output of the compensator at a predetermined ratio. (2) a fuzzy inference device, a compensator comprising a linear or non-linear control section, and a fitness value calculator that calculates a fitness value of the output of the fuzzy inference machine and a fitness value of the output of the compensator from fuzzy operation contents; , an adder that adds the two matching values.