JPH02273804A - Parameter control method for pid controller - Google Patents

Abstract

PURPOSE:To attain the control of a PID parameter during the normal PID control and to prevent the disturbance of a controlled variable by obtaining a correction coefficiency of the PID parameter based on the error value between the characteristic of a control model and that of a control subject. CONSTITUTION:A PID control part 5 controls the manipulated variable to a control subject 7 based on the difference between the target value and the controlled variable and a set PID parameter. A model calculating part 1 obtains the characteristic of a model to be controlled from the PID parameter set at the part 5 so as to secure an optimum PID parameter. At the same time, a characteristic measuring part 2 reads the controlled variable of the subject 7 and measures the characteristic of the subject 7 based on the actual present control waveform to give the characteristic to a characteristic comparison part 3. The part 3 calculates the error value between the actual characteristic of the subject 7 and the characteristic of the model to be controlled. Then a fuzzy inference part 4 obtains a new PID parameter based on the membership function set previously and a rule and changes the parameter of the part 5. Thus the PID parameter is automatically controlled.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application This invention provides 3J! Parameters of the PID control device according to the invention (b1 Overview of the invention)
The regulation method optimizes the parameters of a PID control device that controls a certain control system for the controlled object and automatically adjusts them.

(C1 Conventional technology The method of adjusting PID parameters in a conventional PID control device is broadly divided into auto-tuning method and self-tuning method. As the auto-tuning method, there is a limit cycle method and a step response measurement method. There is a method of determining the P!D parameter.Also, as a self-tuning method, there is a method of adjusting the PID parameter to an optimum value by using the least squares method or pattern recognition of a step response waveform.

ld) Problems to be Solved by the Invention However, in general, according to the auto-tuning method, it is necessary to apply a manipulated variable for identification to the controlled object.
The control amount is disturbed and it is not possible to tune the controlled object while actually performing PID fbllJ. Furthermore, it was difficult to make fine adjustments after tuning, and if the characteristics of the controlled object changed, it was necessary to tune again.

On the other hand, according to the conventional so-called self-tuning method, fine adjustment after tuning is possible, but initial adjustment is difficult. Therefore, usually initial adjustment is performed using the auto-tuning method described above, and while performing actual PID control,
It was necessary to perform the above-mentioned self-tuning or repeat the target value response many times to make fine adjustments. Further, the processing program for parameter adjustment is complicated and requires a large capacity memory.

The purpose of this invention is basically a self-tuning method,
It is also an object of the present invention to provide a parameter adjustment method for a PID control device that solves the various problems described above (Means for Solving the Problems) In this method, the characteristics of a control model that is optimal with the PID parameters are determined from the current PID parameters, the characteristics of the actual controlled object are measured, the characteristics of the control model are compared with the characteristics of the actual controlled object, and both are determined. The feature is that the next PID parameter is determined from the amount of deviation in characteristics.

(f1 action FIG. 1 shows an example of the configuration of the present invention as a block diagram.

In FIG. 1, reference numeral 6 denotes a PID parameter adjustment device for adjusting the PID parameters of the PID control section 5. If the controlled object 7 is, for example, a control system that maintains the liquid temperature at a predetermined value, its input is the heating amount of a heater, and its output is the output of a temperature sensor or the like that measures the liquid temperature. The PID control unit 5 controls the amount of operation for the controlled object 7 based on the difference (deviation) between the target value and control (2) and the set PID parameter. From the PID parameters set in the PID control unit 5, that is, the proportional gauge Kp, the integral time Tl, and the differential time Td, the model calculation unit 1 determines the characteristics of the model of the controlled object such that the PID parameters are optimal. As for the characteristics of the model, for example, the dead time and the maximum slope are calculated backward from the PID parameters using, for example, the Ziegler-Nichols PID parameter design formula shown below.

Kp=1.2/RL Ti-2L Td=L/2 where R is the maximum slope of the step response and L is the dead time. Also, calculate the time constant etc. backwards using other formulas.

This means that "what kind of control object the current PID parameters are optimal for" is being sought. Furthermore, the model of the controlled object obtained in this way is used as the current PI.
For example, overshoot, settling time, damping ratio, etc. are determined as characteristics when controlled using the D parameter, and are provided to the characteristic comparing section 3.

On the other hand, the characteristic measurement unit 2
c is read and the characteristics of the controlled object such as overshoot, establishment time, fIi damping ratio, etc. are measured from the current actual control waveform and are provided to the characteristic comparing section 3. The step response waveform of the control object 7 is as shown in FIG. 2, for example. The characteristic measuring section 2 detects overshoot 1 from such a response waveform.
``A + Determine the established timetable and damping ratio b/a.

The characteristic comparison unit 3 calculates the deviation (eg, ratio) between the characteristics of the actual controlled object and the characteristics of the controlled object model. The fuzzy inference unit 4 performs fuzzy inference from the deviation amount of the characteristics based on predetermined membership functions and rules, calculates a change (correction coefficient) of the PID parameter, and multiplies the current PID parameter by the correction coefficient to create a new PID parameter. PID
Obtain the parameters and change the parameters of the PAD control unit 5. In the above manner, the PID parameters are automatically set to 1! adjustments will be made.

(g) Embodiment FIG. 3 shows a block diagram of a PID parameter 31 device which is an embodiment of the present invention.

In FIG. 3, the CPU 13 controls the entire PrD parameter adjustment device. A processing program for the CPU 13 is written in the ROM 14 in advance. The RAM 15 is used as a working area for storing current PID parameters, sampling data, and the like. A sample and hold circuit 10 samples and holds the control I, and an AD converter 1) converts the sampling signal into a digital signal. CPU13 is I10 boat 12
sample the control amount at the necessary timing and read the data. DA converter 17,
18.19 are Kp respectively.

This is a circuit that converts Ti and Td parameters from C to D.
By outputting these PID parameter data via the I10 boat 16, the PU 13 sets the PID parameters of the PID control unit 5 shown in FIG.

FIG. 4 shows the configuration of the main part of the storage area of the RAM 15. In the figure, Ml is the current PID parameter, and is an area for storing Kp (proportional gain), Tl (ff1 minute time), and Td (differential time). M2 is PID parameter change r1 (correction coefficient), and the next PID parameter is (Kp・Ap), (T*-
Ai), (Td, BooAd). M
3 is an area for storing characteristics when the controlled object model is controlled by the current PID parameters, O8 is an overshoot, TS is a settling time, and DR is a damping ratio. M4 is an area for storing the actual characteristics of the controlled object obtained by measuring the control amount, O3r is the overshoot, TSr is the settling time, and DRr is the damping ratio. Further, M5 is an area for storing sampling data of control (2).

Next, the processing procedure of the PID parameter adjustment device is shown in FIG.

When performing a new target value response or when the characteristics of the controlled object change, first the current PID parameters Kp, Ti, Td
Then, the dead time, maximum slope, and time constant are determined as the characteristics of the controlled object model for which these parameters are most suitable by back calculation using the Ziegler-Nichols PID parameter design formula, etc., and then the controlled object model is controlled using the current PID parameter makeup. Overshoot O8 as a characteristic when
, find the settling time TS and damping ratio DR (nl-n3
) -Vt, target value response control is performed on the actual controlled object, the response waveform is sampled, and the overshoot O3r is determined as the characteristic of the controlled object as shown in Figure 2.
<=a), settling time TSr (-t) and damping ratio DR
Find r(-b/:a) (n4). Furthermore, the ratio of the characteristic parameter of the actual controlled object to the characteristic parameter of the model already determined is determined as follows: ratio of overshoot=O3r10S O3 time=TSr/TS ratio of damping ratio-DRr/DR (n5). Next, fuzzy inference (described later) is performed from these characteristic ratios to obtain PID parameter correction coefficients (Ap, At, Ad) (n6), and the next PID parameter is determined by multiplying the current PID parameter by the parameter correction coefficient. This is calculated, stored as the current PID parameter, and output to the PfD control unit 5 (n7→n8).

Next, a method of calculating the correction coefficient of the PID parameter will be described.

In this embodiment, fuzzy inference is performed using the ratio of the characteristics of the actual eight controlled objects to the characteristics of the controlled object model as an input variable to obtain the PID and parameter correction coefficients.

6 (A) to (C) are membership functions of input variables, (A) is a membership function of the overshoot ratio (hereinafter referred to as S1), and (B) is a membership function of the damping ratio (hereinafter referred to as S1). (C) shows the membership function of the settling time ratio (hereinafter referred to as S3).

The label and scale of each membership function are as shown in the figure, with the center of label ZR being 1, 1 or more being positive, and less than 1 being negative.

FIG. 7 shows the membership function of the modification coefficients of the PID parameter, which is the output variable, and shows the three modification coefficients A p l
This is common to A i+ A d.

FIG. 8 shows fuzzy rules using the various membership functions shown above. Here, 31°S2, 33 are variables of the antecedent part, Ap, At, Ad are variables of the consequent part,
0 where the contents of the matrix represent the labels of each consequent
For example, if the overshoot ratio S1 is PB, the damping ratio ratio S2 is PM, and the settling time ratio S3 is PB, then A
p (correction coefficient of Kp) is increased to PB, At
(T1 correction coefficient) is reduced to NB, Ad
According to such a membership function and fuzzy rules that make (modification coefficient of Td) ZR, that is, approximately the same, PI
FIG. 9 shows the procedure for determining the correction coefficient of the D parameter.

First, a correction coefficient to be determined is set (nlO). This will be Ap in subsequent processing.

Set which of At and Ad is to be determined.

Next, for each characteristic ratio SL, S2, and 33, the member ratio for each label is calculated (n1)). Subsequently, the rule data is read out, and an antecedent logical AND operation is performed by finding the minimum value among the membership values corresponding to the antecedent among the membership values of each characteristic ratio already determined (n12-n13). Next, the consequent logical product operation is performed by truncating the membership function of the consequent by the minimum membership value of the antecedent (n
14), perform a logical sum operation on the inference results by finding the maximum value among the trapezoidal parts cut off in this way (n1
5) After executing the processes of n12 to n15 for all rules regarding correction coefficients to be determined, determine the center of gravity position of the ORed region (n 1 G−n 1 ?),
This center of gravity value is the final value of the correction coefficient. Vi, similar processing is performed for the other correction coefficients, and three correction coefficients Ap, At, and Ad are obtained by performing the processes n1) to n17 for the three correction coefficients.

In the embodiment, the PID parameter adjustment device is configured by a microprocessor system, and the PID parameter adjustment device is configured by a microprocessor system.
I configured it to give an ID parameter, but this PID
It is also possible to configure the entire control section and PID parameter adjustment device by a microprocessor system. Further, in the embodiment, an example is shown in which fuzzy inference is performed by digital arithmetic processing, but it is also possible to perform this by a fuzzy inference circuit. Furthermore, processing programs and circuits that perform other functional operations other than fuzzy inference based on the amount of deviation between the characteristics of the control model and the characteristics of the actual controlled object,
It is also possible to determine correction coefficients for PID parameters.

(!1) Effect of invention According to this invention, the following effects are achieved.

(1) Since the adjustment can be made while performing normal PID control, there is no need to add a manipulated variable for identification (and the controlled variable will not be disturbed).

(2) During initial adjustment (when the current PID parameters do not match the controlled object), during '&1m adjustment (when the current PrD parameters are approximately the optimal parameters), or when the characteristics of the controlled object change It can be applied to any case, and there is no need to perform different controls depending on each state, and the processing program does not become complicated.

(3) Approximately optimal PID parameters can be obtained with a single target value response, such as during the first adjustment, and there is no need to repeat the target value response many times for adjustment.

(Since there is no need to constantly perform complicated calculations like the 41m fractional square method, and there is no need for a large memory capacity like the pattern L'12 method, the memory capacity M and program capacity do not increase. .

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an example of the configuration of the present invention. FIG. 2 is a diagram showing an instantaneous target value response waveform. FIG. 3 shows the PID parameters 1!1 which is an embodiment of this invention.
FIG. 4 is a block diagram of the control device showing six configurations of the main parts of the RAM in the device. FIG. 5 is a flowchart showing the processing procedure of the device. Figure 6(A)-(C
) and FIG. 7 are diagrams showing various membership functions, and FIG. 8 is a diagram showing fuzzy rules. Figure 9 shows the P! of the same device. 3 is a flowchart showing a procedure for calculating a D parameter correction coefficient.

Claims (1)

[Claims] (1) In a method of PID controlling a controlled object based on the deviation of the controlled variable from the target value, the characteristics of the control model that is optimal with the PID parameters are determined from the current PID parameters, and the characteristics of the actual controlled object are measured. A parameter adjustment method for a PID control device, characterized in that the characteristics of the control model are compared with the characteristics of the actual controlled object, and the next PID parameter is determined from the amount of deviation between the two characteristics.