JPH0194401A - Fuzzy operation estimation system for auto-tuning controller - Google Patents

Abstract

PURPOSE:To quickly attain the coping even when the inclination of a divergence is detected while the tuning to grasp correctly the control system condition can be executed by dividing a control rule is divided into two types of a priority instruction and a general instruction and issuing the priority instruction depending on conditions and the general instruction for a constant time interval. CONSTITUTION:An auto-tuning controller divides a control rule into the priority instruction and the general instruction at the time of the output of the control instruction, the priority instruction is issued immediately when the control conditions are established, and the general instruction is issued in accordance with the condition of the system during then after the preceding control instruction is issued and the instruction time interval passes. Thus, at the priority instruction, for example, when the inclination of the divergence is detected, the coping can be quickly coped, and at the general instruction, the temporary noise influence is avoided, the control is executed so as to execute the tuning to correctly grasp the condition of the control system.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a fuzzy operation estimation method for an autotuning controller that automatically adjusts control parameters based on fuzzy inference in a controller used for example in process control. It is something.

[Conventional technology]

In general, the concept of fuzzy, which can be seen in terms such as fuzzy estimation calculation and fuzzy adjustment, is a relatively new concept that appeared around 1966, so it will be briefly explained below to the extent necessary for understanding the present invention.

゛ Fuzzy is an English word that means ``ambiguous.'' So-called “fuzzy mathematics” such as fuzzy set theory, fuzzy logic, and fuzzy algebra
has appeared in recent years, and the control performed by applying this fuzzy theory is called fuzzy control, and it is said to be comparable to conventional process control.

The purpose of ambiguous control is to use computers to control complex processes that cannot be modeled mathematically, but which a single person (operator) would be able to successfully implement based on vague knowledge of process behavior and control experience. This is where I try to reproduce it.

Therefore, as a method for this ambiguous control, in addition to the process measurement values, the results obtained by the operator's subjective and intuitive judgment of the process situation (in other words,
The results of judgments such as small, medium, high, or low are called ambiguous variables. ) to perform inference (estimation calculation) according to predetermined ambiguous rules, and use the result as a control output. Here, regarding the ambiguous variable, a membership function is determined in advance that represents the proportion of the number of people who judge that the same process situation is ``large'', and this membership function is used to determine the ambiguous variable. A numerical value (this numerical value will be referred to as a process status value here) is calculated.

Since the present invention is not aimed at ambiguous control itself, further explanation will be omitted.
VoL, 18. Na2. PL50-160 papers, “Fuzzy sets and their application to logic”, also “Total i1+ff
Control J VoL, 22. Please refer to the paper on pages 84-86, "Ambiguous Control," etc. In this specification, we believe that the term "ambiguous" may still cause misunderstandings and is not appropriate, so we will use the term "ambiguous" in this specification. This will be explained using the term "fuzzy".

Next, let's move on to the explanation of the drawings.

FIG. 6 is a diagram showing fuzzy inference in the simulated fuzzy estimation calculation method of the conventional fuzzy adjustment device disclosed in Japanese Patent Application Laid-Open No. 60-204002. In the figure, 6-1 indicates that the deviation is NS (Negative Small
) membership function, 6-2 is the time change of deviation P
Membership function of S (Positive Small).

6-3 is the membership function where the amount of change in the manipulated variable is PS,
6-4 is a membership function in which the deviation is ZO (Zero), 6-5 is a membership function in which the time change of deviation is ZO, 6-6 is a membership function in which the amount of change in the manipulated variable is ZO, 6-7 is a synthesis of 6-3 and 6-6 above.

Next, a fuzzy inference synthesis method using a conventional fuzzy estimation calculation method will be explained. First, control rules generally take the following form.

IF P THEN Q (If P, then Q.) ... (1) P consists of multiple propositions for conditions and is called an antecedent proposition, and Q
indicates a proposition for the output, and is called a consequent proposition.

Therefore, as shown in FIG. 6, rule 4, IFe=Nsa
ndle=Ps-+711u=Z○, and IF of rule 5
In a system consisting of two control rules e=ZOandΔe=Zo−+711u=Z○, the input eatden
In this case, the amount of change du0 in the manipulated variable is determined. (However, NS is negative small.

PS stands for positive small and zO stands for zero. ) In Rule 4, the fuzzy set NS is satisfied to degree 16
-1, the fuzzy set PS is satisfied to a degree of 0.86-2, the transformation ω1 that satisfies the antecedent proposition of rule 4 is 2
The smaller value of the degree of satisfaction of two fuzzy sets is 0
．． 86-3° The degree to which the antecedent proposition of Rule 5 is satisfied ω2 similarly becomes 0.56-6° Therefore, the fuzzy set representing the amount of change ΔU in the manipulated variable is calculated for each control rule. The degree to which the antecedent proposition is satisfied ω
1. 6-3゜6-6, which is reduced according to ω2, gives the evaluation value obtained as above to the consequent proposition, and calculates the weighted logical sum of these two fuzzy sets 6-3゜6-6. The amount of change in the manipulated variable Auo is determined by averaging, and the amount of change in the manipulated variable I! is calculated for each sampling period. 1 Iua is calculated and output at the same timing.

U here. =X, then the amount of change in the manipulated variable, ! juo is expressed as (ρ(x): evaluation value of the logical sum of 6-3.6-6).

[Problem that the invention seeks to solve]

The fuzzy inference of the conventional fuzzy estimation calculation method is performed as described above, so if you try to apply it as is to the fuzzy estimation calculation method of the autotuning controller, it will be difficult to respond to temporal noise. There have been problems such as incorrect tuning, inability to accurately grasp the state of the control system, and frequent output of manipulated variables, resulting in unstable tuning.

This invention was made in order to solve the above-mentioned problems.- It is possible to perform tuning that accurately grasps the state of the control system without being affected by temporal noise, and it also makes it possible to detect the tendency of divergence in the control system. The purpose of this study is to obtain a fuzzy calculation estimation method for an autotuning controller that can be used to quickly deal with such situations.

[Means for solving problems]

The fuzzy calculation estimation method for the autotuning controller according to the present invention provides rules for control commands divided into priority commands and general commands, and for the priority commands, the control commands can be issued immediately as soon as the control conditions are established. ,
General commands are issued after a command time interval has elapsed since the previous control command was issued, depending on the system status during that time.

[Effect]

The auto-tuning controller according to the present invention divides the control rules into priority commands and general commands when outputting control commands, so that the priority commands are issued immediately as soon as the control conditions are established, and the general commands are issued immediately after the control conditions are established. By making it possible to issue commands after a command time interval has elapsed since the previous control command was issued, depending on the state of the system during that time, priority commands can - The general command avoids the influence of temporary noise and performs tuning that accurately grasps the state of the control system.

[Embodiments of the invention]

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

In Figure 1, 1-1 is the membership function for "the control system is oscillatory", and a is expressed by the formula IP3-P21, which is the difference between the current deviation peak and the previous deviation peak. It is the absolute value of the difference, and the larger the absolute value a of the difference, the greater the tendency for the control system to be oscillatory.
1-2 is the membership function of 1'"poorattenuation", b is expressed by the formula IP3-P21/IP2-Pi I, and the absolute value a of the current difference is set as the absolute value a of the previous difference. The larger the ratio of the absolute value of the difference between the previous time and this time, the worse the damping performance is.1-3
is the membership function of gain KcDOWN.

Next, in Fig. 2, 2-1 indicates the "first deviation peak," 2-2 indicates the "second deviation peak," 2-3 indicates the "third deviation peak," and P3 indicates the "first deviation peak." The peak of the deviation closest to the current time is shown, P2 is the previous deviation peak, and Pl is the deviation peak before the previous time. Also, in Figure 3, 3-
1 is a membership function for which "overshoot S is small", and S indicates the 'overshoot amount'.

3-2 is a membership function whose absolute value of error is large, 3-3 is a time average of the degree to which control rule 2 is satisfied, and 3-4 is a member whose overshoot S is small. The ship function, 3-5, is 'absolute value of error lel is small'
The membership function 3-6 is the time average of the degree to which control rule 3 is satisfied.

Furthermore, in FIG. 4, 4-1 is 'gain KcUP'
The membership function of 4-2 is.

Membership function of “gain KcKEEP”, 4-
a +i This is a synthesis of 4-01 and 4-2 above. here,
Each rule is described below. Rule 1 is that the control system is oscillatory and the damping properties are not good.→C is down for the gain. Rule 2 is that the overshoot S is small and the absolute value of the error is l.
e+ is large, → gain Kc is UP, rule 3 is that overshoot S is small and absolute value of error 1el is small, →
Gain Kc@Keep.

Also, in Fig. 5, 5-1 is an auto-tuning
A controller, 5-2 is a fuzzy calculation estimation unit according to the present invention, 5-3 is a PID controller, 5-4 is a controlled object, 5-5 is a target value SV, 5-6 is the Sv and the controlled object 5-4 5-7 is the output of the PID controller or -3, the manipulated variable CV, 5-8 is the control object 5
-4 output, control amount PV, 5-9 to 5-11 are the above-mentioned PI
The fuzzy operation estimation 5- for the D controller 5-3
2, 5-9 is the gain tuning amount 4, 5-10 is the integration time tuning amount ΔT, 5-1
1 is the tuning amount ΔTo of the differential time.

Next, the operation of the autotuning controller according to the present invention will be explained. First, as shown in FIGS. 1 and 3, Rule 1. Rule 2. In a system consisting of the three control rules of Rule 3, when inputs (aot bo) and (s, , Ie 1.) are obtained, the gain operation amount ΔKc is determined. In the present invention, the control rules are a priority command (Rule 1) and a general command (Rule 2,
3) It is divided into two parts.

As shown in Figure 1, in Rule 1, the fuzzy set ``The control system is oscillatory'' is satisfied to a degree of 0.5, 1-1, and the fuzzy set ``The damping properties are poor'' is satisfied to a degree of 0.75. 1-2, the degree G1 to which the antecedent proposition of Rule 1 is satisfied is 0.5, which is the smaller value of the degrees to which the two fuzzy sets are satisfied.

Since Rule 1 is a priority instruction, it is output immediately as soon as G1 is established as a positive value.

As shown in Figure 3, in Rule 2, the fuzzy set "overshoot S is small" is satisfied to a degree of 0.5, and 3-
1. The absolute value of the fuzzy set error is large” is 0.3
3-2, and the degree G2 that the antecedent proposition of Rule 2 is satisfied is 0.3, which is the smaller value of the degrees that the two fuzzy sets are satisfied. If the time rate of this instantaneous value G2 is averaged and the degree of establishment of rule 2 used for tuning is GRAD2, then to is the time when an instruction according to one of the rules is issued, and CMDTIME is the instruction time interval.

Similarly for rule 3, 03 becomes 0.5 and GRA
D3 gives the evaluation value obtained in the above manner obtained by equation (3) to the consequent proposition, and calculates the change in the manipulated variable 7! Find lKc and find the time t, + after the command time interval has elapsed since the command according to one of the rules was issued.
Issue as a tuning command to CMDTIME.

As shown in Fig. 5, SV, 5-5, and output 5 of the controlled object.
The deviation EV, 5-6 from -8 becomes the input to the PID=+controller 5-3, and its output is the manipulated variable CV, 5-7.

The manipulated variable CV becomes an input to the controlled object 5-4.

Its output is the control amount Pv, 5-8. The control amount Pv is used to obtain the deviation EV from the new target value Sv,
The deviation EV becomes an input to the PID controller 5-3 again. Now, the fuzzy calculation estimating section 5-2 according to the present invention operates to calculate the target value SV, 5-5. Deviation EV, 5-6
．． Using the operation icv, 5-7 and the control amount Pv, 5-8 as input, a fuzzy calculation estimation is performed in which priority instructions and general instructions are set separately as described above.

Although the above embodiment has been described with respect to proportional gain tuning, it can also be applied to integration time and differential time tuning, and can be applied not only to velocity type but also to position type. Furthermore,
In this example, a case has been described in which there is one priority instruction and two other instructions, but the number of instructions may be any number.

Furthermore, the same effects as in the above embodiments can be obtained not only in tuning but also in control.

〔Effect of the invention〕

As described above, according to the present invention, control rules are divided into two types, priority commands and general commands, and priority commands are issued immediately as soon as conditions are met, and general commands are Since the commands are issued at each time interval according to the state of the system during that time, general commands are not affected by temporary noise and can be tuned while accurately grasping the state of the control system. This has the effect of allowing immediate action to be taken even if a trend is detected.

[Brief explanation of the drawing]

FIG. 1 is an explanatory diagram of a priority command output method according to an embodiment of the present invention, FIG. 2 is a waveform diagram showing the peak of deviation, and FIG. 3 is an explanatory diagram of the antecedent proposition of a control rule according to an embodiment of the present invention. FIG. 4 is an explanatory diagram of a method for determining an evaluation value from the evaluation value determined in FIG. 3. FIG. 5 is an autotuning controller that performs fuzzy operation estimation according to the present invention. FIG. 6 is an explanatory diagram showing a conventional instruction composition method. In the figure, 1-1 is a membership function that indicates "the control system is oscillatory", 1-2 is a membership function that indicates "poor damping", 1-3 is a membership function that indicates "gain KcDOWN", and 3 -1 means "overshoot S is small"
3-2 is the membership function indicating “absolute value of error 1
3-3 is the time average of the degree to which control rule 2 is satisfied, 3-4 is the membership function that indicates "overshoot S is small," and 3-5 is the absolute error of ``el''. 3-6 is the time average of the degree to which control rule 3 is satisfied,
4-1 is the membership function of 1′ gain KcUP”,
4-2 is a membership function of "5 gain KcKEEP", 4-3 is a combination of 4-1 and 4-2, 5-1 is an auto-tuning controller, 5-2 is a fuzzy calculation estimator, 5-3 is a PID controller, 5-4 is the controlled object,
5-5 is the target value SV, 5-6 is the deviation EV, 5-7 is the manipulated variable CV, 5-8 is the controlled variable PV, 5-9 is the gain tuning amount JK, and 5-10 is the integration time tuning amount J
Tt, 5-11 is the tuning amount Δ'ro of the differential time.゛Patent applicant: Mitsubishi Electric Corporation Figure 4

Claims (1)

[Claims] In a fuzzy operation estimation method for an auto-tuning controller that converts the control status of a process into a membership function and outputs a control command to the process in order to automatically adjust control parameters based on fuzzy inference, the control command to the process is Rules are set separately for priority commands and general commands, the priority command is set to issue a command immediately when a control condition is satisfied, and the general command is set to issue a command time interval after the previous control command is issued. A fuzzy calculation estimation method for an auto-tuning controller, characterized in that a tuning command is issued according to the status of a control system during that time after a certain amount of time has elapsed.