JPS5983208A - Auto-tuner - Google Patents

Abstract

PURPOSE:To set automatically parameters of a PID controller by integrating input and output signals to an object system and identifying the object system on a basis of integrated values. CONSTITUTION:A sequencer 16 generates pulses 22-25 in accordance with push- button operations of the operator. Integrators 8 and 19 which receive input and output (u) and (c) of an object system 1 integrate them synchronously with the pulse 25. Both integrated values are supplied to an identifier 14 through A/D converters 11 and 12 and a data buffer 13. The identifier 14 identifies parameters of the controlled system 1 synchronously with the pulse 22. When receiving this result, a control parameter calculator 15 outputs PID control parameters synchronously with the pulse 23, Meanwhile, a test signal generator 17 generates a test signal synchronously with the pulse 24. Thus, parameters of the PID controller are set automatically.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an autotuner that automatically tunes control parameters of a feedback controller using P used in process control.

Conventionally, controller parameter tuning has been performed manually in the field by changing parameter setting knobs and observing actual fluctuations in control variables, which has improved the efficiency of instrumentation and stem adjustment work. It was left and right.

This invention aims to eliminate these drawbacks.

In a closed loop state with a controller, add a small step signal to the operating end, integrate the input and output signals to the target system, use the fact that these become step signals to identify the target system, and calculate the result. By uniquely setting parameters from P based on P, you can automatically perform stable auto-tuning in a short time without changing the set values and only changing the output of the target system within the allowable range. The purpose is to do so.

The autotuner according to the present invention will be described below using examples. FIG. 1 shows a situation in which the autotuner according to the invention is used. In the figure, (1) is the controlled object, and its transfer function G(8) can be approximated by either a first-order lag including dead time or a second-order lag including dead time. Here, K is called a gain, T is a time constant, and A and B are called secondary system parameters. It is known that many controlled objects are expressed by Equation (11) or Equation (2).In order to maintain the output C of the object system i11 at the set value r, feedback control is performed using the pressure controller (2). (3) is a vacuum reducer and is normally connected to controllers (2) and 1.
It has become a body. (4) is an auto tuner, and the test signal d from this is added to the set value r by an adder (61) to become the effective set value p. q is the control deviation and U is the manipulated variable. ) is the transfer function (
) (sl is G(B)=Kp (1+% 1'r,, B )
(Given by 31. Control parameters Kp, TI,
TD is set by the auto tuner (4).

(5) is a signal requesting the start of the auto tuner and is set by the operator.

FIG. 2 is a block diagram of the configuration of an autotuner according to the present invention. In the figure, Ol) and α are A/D converters corresponding to the input and output to the target yarn ill, (13 is a data buffer, 14 is an identifier, α is a control parameter calculator,
(161 is a sequencer, αT is a test signal generator.

Also, c! 2 is an identification pulse, and 2 is a parameter pulse.

is the test signal pulse.

The operation of each component block will be explained below, and it will be described how the function of the autotuner is achieved.

First, the operation of the integrator and A/D converter 011 will be explained. When the integral pulse (to) described later becomes ON, the lees separator US resets the integral value and assumes the target system (110 manual input integrals. This signal U (t) is A every sampling period T.
/D converter full converts the A/D signal into a digital signal Un. That is, Um==U (mTs) +51
It is.

In the same way as in the case of the target string (output CKpAI of 11, input uO), the integral value is obtained by the integrator aS, and the digital signal Om=0 (mTB) (7) Become. However, it is assumed that the A/D converters αB and nz operate synchronously.

That is, as shown in FIG. 4, the integral signals U and C of the target yarn (110 person output signals U and C) are step signals, which are identified as input and output to the virtual target system +11. The background that allows this can be stated as follows.The target system (transfer function G(8) of 11 and its human output u(t),
During c (t), L (c(t) ) = G(s)
There is a relationship L[u(tl) -.

At this time, from equation (41) and equation (6), L (U(t
l j = ' L (u(tl) --
−+91L (c(t>)=−L−L t C(tl)
...((- becomes. Therefore, No. (8)
Using Equations 2 to 10, it is shown that identification is also possible using the precision of inputs and outputs U and C in object system (11).

Next, the operation of the sequencer (161) will be explained. When the tuner start signal (5) turns on, the test signal pulse a41 is turned on as shown in FIG. , and then an identifier (+41
Turn on the parameter pulse (c) after a time that allows sufficient calculation time. Furthermore, the integral pulse (c) is turned on at the same time as the test signal pulse axis, and 1iiPF is set in the middle of the time when the identification pulse node parameter pulse (c) is turned on.

Let's talk about the settling time TQ. When the test signal d is output, the target thread (110 input/output U and C change as shown in Figure 8144. Signals U and 0, which are divided by the integrator and ol, are shown in the figure. As shown, the signal becomes a step-like signal.The settling time 'rq is as small as possible when the signals U and C are considered to have settled sufficiently.Then, the number of sample points N for the sampling period Tθ is expressed as N=(M). ] T6 It is defined as °°°0'. However, [・] is the symbol of Kakara.

Next, the operation of the test signal generator α9 will be explained.

When the test signal pulse turns ON, an analog signal is generated that rises at a constant slope from the 0 level to the sH level after time TB as shown in FIG. 4. The time TB and level SH can be set by an external operator.

Next, the operation of the data buffer 03 will be explained.

The digital signals Um and On, which are updated every sampling period T8, are stored from the present to the past (N-1) Ts while being updated every sampling period T8. Let m −Ts be the current time.

Xk=Um-N+1+k...-f13
7k” (3m-N-1-1-1-k -(
141 (""0 #1 e...e N*), the value of the Shibara fur memory is updated.

That is, lN-1 is CI, that is, the current value, and x.

is Cm-N+1, which represents (N-1) Ts past values. The output (2) of the data buffer α3 outputs the vector (XO*
is a vector (YO@Yl m ”・+ YN-1)
It represents.

Next, the operation of the identifier α will be explained. Identification pulse @ is ON
Then, perform the calculation described below.

Data buffer 0 contains data related to the inputs of the target system (11)
is stored.

The identifier (141 uses these values to estimate T, A, B, and L for the parameters of the target system +11. Integer MB TB MB =, (,,)...o9
Let's define.

First, the next ↑'B, TB, yB, "j",
Calculate.

, 52. B=1 〒1Xk...u
tiMB](=[l ↑8=YMS-1.N-2-ar+MB k=0 Since these give the initial and final setting values of the input and output, the estimated value for the gain of the target system is is given by

Next, the signal Xk, 7k is defined from (k=Os 1 *..., 2N-1)K. xk and 7 are called value inversion connection signals of Xk and Yk, respectively.

Next, the estimated value of the power spectrum ↑X(- and pesos (ω) and the power transmission ratio total (ω
)of? (ω)=? y(ω)/乍X(ω)...(
c). Additionally, calculate the function. For the transfer function G(s) of the object system (1).

G1(1ω)12-↑(ω)...(27
1, so the object system (11 forces Z i
(If it is a first-order delay that includes the dead time of Equation 11, then the +11th
From the equation, the equation (a), and the gradient equation, we get the total (store-T2 ``'@.

Similarly, if the target system t1) is a second-order lag including the (2) 20th dead time, then Q(m'=i (A2-2 B >+ω B ・
...0! l [becomes. Next, let the angular frequency resolution be C0 for the observation time TQ, and define ω0=π/M TQ . Here, M is an integer of about 10 to 30. Furthermore, MT=3・M... is selected like a circle. And/131=仝(lωo) C1=1.2.・yT＊i
Φ2M)6. Especially 12M = '(?zy-t +'¥'2M+1
) Defined in G31. In this way, C1=1.2. ...e MT) is obtained.

Considering equation (2), if the target system (1) is given by equation (11), then the estimated value of the time constant T is given by:

Also, if we consider the No. 1 Kan equation and assume that the target system (1) is given by equation (2), the estimated values of A and B are given by: Here, if β〉0.

- Contains 1,000 2β carbon dioxide (9). The combinations show α and β that minimize the following quadratic equation.

Next, let ek and fk be the outputs of maki that add input Xk (k=Os 1 e..., 2N) to the system ↑ 0° (sl = = -; 1 - 01) that does not include dead time. is determined by numerical integration.MQ is determined as an integer of about 10 and is determined sequentially by Euler's method as follows.

If we consider that the target system (1) is given by equation (1), then ek, Q ” ek eg : Q
(4118 eksj+t=ek,j+MQT (K Xk e
k, j) f42 (j=0, 1,..., Mq
-1) ek-1-t = ek, Mq -1(431(
1(==Q s t -...), e)( is obtained.

Furthermore, if we consider that the target system +1) is given by equation (2), then (k, 0=(k, fk, 0=fk, (0=O,
fO=OA:! FO, B km (j=o, 1...., Mql) (k-1-s=(k, Mq-1f'ri fk+t = f
k, Mq-1 (
481 (k=0.1...) 1 Calculate by K. t, ek and f at k〈0
The kO value is set to 0.

Then, the following two equations are defined. In other words, this is a formula that expresses the error in the output of the system when there is no dead time and when there is dead time.

(j””Om 1 #”’m kMAX) where k
MAX is an integer that is expected to be larger than the value of kMAxT8 plus the dead time LJ:. Maximize j and Jj
Let them be Jy and js, respectively. At this time, JF≦Jj8
If so, it is assumed that the target system is actually a first-order system, and if not, it is assumed to be a second-order system.

and This is the estimated value of the dead time.

The output of the identifier αa is given as follows. That is, xB=f (52/^^LE = L
551.

Next, the operation of the control parameter calculator (3) will be explained.

When the parameter pulse (c) turns ON, the following operation is performed.

If the transfer function of the controller (2) is Gc (sl), then the closed loop transfer function H(8) is cG H(8) = 10G.

Then, the controller Gc(s) becomes 581 Ga(s) :KL< 1+a)(1+〒1) when the target system 0) is equation (1),
The target system +11 becomes the th (in the case of the 21st equation), where a=λL is the older sister. That is, both the th side equation and the 5th!J equation become controllers than P.

It is known that the right side of the 15th η equation shows a step-like rise from overbraking to a state where overshoot occurs depending on the value of X, and if a is constant, it shows the same shape: >, a=o, szs, it is known that the closed-loop step response produces an overshoot of about 5%. Therefore, the control parameter calculator (3) outputs control parameters as follows. However, a is a=0.825
Use day.

When trench BB=00 (primary system).

'rI=TH13 TD=OII)41 and B, for ~0 (second-order system).

'l' engineering=AB
Output as 6) E TI) = □ E and rewrite the parameters in controller (21).

With such a configuration, it is possible to easily estimate (identify) the characteristics of the target system, and to set parameters using a simple and unique P, thereby achieving a function as an autotuner.

As described above, the present invention makes it possible to automatically perform parameter tuning operations in instrumentation equipment, which not only makes it easier to adjust the instrumentation, but also eliminates variations due to the individuality of the operator. This means that highly reliable instrumentation can be configured.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the environment in which the autotuner is used, and FIG. 2 is a block diagram of the autotuner configuration. FIG. 3 is a diagram showing the operation of the sequencer, and FIG. 4 is a diagram showing changes in test signals and target system input/output signals. In the figure, (1) is the target system, (2) is the controller, and (3)
is a subtracter, (4) is an auto tuner, +III, f
ly is an A/Df converter, 0i is a data buffer, α4) is an identifier, 09 is a control parameter calculator, (+61 is a sequencer, and 071 is a test signal generator. In Figure 1, the identity symbol indicates the same or The relevant parts are shown. Agent Shin Kuzuno - Figure 1 Figure 2 - Figure 3

Claims (1)

[Claims] In PID feedback control systems used in various processes, the first . Second.
a sequencer that generates third and fourth pulses, first and second integrators that integrate the input and output of the controlled system in synchronization with the pulses of set 1, and first and second integrators; first and second A/D converters that convert the outputs of the converters into digital signals; a data buffer that stores and updates these two digital signals every conversion cycle; a control balata calculator that outputs a control parameter from P from the estimated controlled object parameter in synchronization with the third pulse; 9. A test signal generator that outputs a step-like test signal to be applied to the operating end in synchronization.9. It is characterized by identifying the controlled object without changing the set value and by making a slight change to the operating end with almost no effect on the control amount, and thereby automatically setting the parameters of the controller from P. Auto tuner.