JPH02183302A - Fuzzy control method - Google Patents

Abstract

PURPOSE:To simplify the constitution and adjustment work of a data holder, etc., by identifying the characteristic of a process and changing the sampling cycle of a fuzzy controller according to the result of this identification. CONSTITUTION:In a control unit 1, deviation and a control rule to a change rate in a fuzzy controller 108 are fixed and a membership function is also fixed. As the function of a DELTAt and K change function 109, a sampling period DELTAt is increased to the increase of a dead time L and a controller gain K is decreased to the increase of a process gain G. Then, to the increase of a read time T, the DELTAt and K are coupled and the increase and decrease is adjusted. Based on the identified result of an estimating means 110 to process characteristics (L, T and G), the t and K change function 109 is driven. Thus, the adjustment is simplified and the controllability of adaptive control is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention provides a method for applying the manipulated output obtained by performing fuzzy calculations at a predetermined sampling period to the deviation between a measured value and a target value of a process to This paper relates to improvements in fuzzy control to provide for.

(Prior Art) An example of a conventional method for determining a membership function in a fuzzy controller for directly determining an operation output will be explained based on FIGS. 7 to 9.

First, the following prerequisites are assumed: ■The control rules are known, ■Actual operational input/output data by veteran operators are known, and ■The membership function is unknown.

Here, the known control rule is, for example, the control rule (1
)...If the hot water temperature inside the can is high, reduce the heating steam and increase the cooling water.

Control rule (2): If the temperature inside the furnace is low, increase heating steam and decrease cooling water.

An example of a membership function that reflects this control rule is
The seventh example is based on the case where a general trapezoid is used as the shape of the function.
As shown in the figure. There are four types: (A) is low, (B) is high, (C) is decreasing, and (D) is increasing. Points a, b, c, d,
P+ q, r, and s points are unknown parameters.

On the other hand, the fact that the actual operational input/output data by experienced operators is known means that, for example, if the temperature inside the can is
When the steam supply amount is y, ((x + 3'
), (x2.y2), ..., (x,,y,>1).

The fuzzy control calculation when the fuzzy calculation is executed on the actual measurement value X· on such a premise will be explained with reference to FIG.

(A) is a computation of control rule (1), and represents a computation in which the manipulated variable is determined by the diagonal line based on the membership function ``decrease'' based on the membership function ``high'' at the actually measured value X.

Similarly, (B) is a calculation of control rule (2), and represents a calculation in which the manipulated variable is obtained by ``increasing'' the number of memberships based on the membership function ``low'' at the actual measurement value X 2 with diagonal lines.

FIG. 9 is an explanatory diagram for calculating the manipulated variable output Y, which is a weighted average of the manipulated variables calculated according to control rules (1) and (2), and the output value Y is as follows: Y.=Y.<x, a, b, c, d, p, q.

r, s) becomes.

Usually, the shape (parameters) of the membership function is determined by repeating adjustment by trial and error so that the fuzzy calculation value Y and the actual operational output data yi are as close as possible.

On the other hand, calculation methods for automatically determining the shape of membership functions are also known. That is, considering the following evaluation function, parameters (a, b, c, d, p, q, r, s) are determined by a nonlinear optimization method so that the evaluation function J has a minimum value. This method requires a considerable amount of time even on a large-scale computer as the number of parameters increases.

This method of parameter determination using the nonlinear optimization method is publicly known, and is disclosed, for example, in "Fuzzy Control JP 147" by Michio Kanno, published by Nikkan Kogyo Shimbun in 1988.

Next, an example of applying fuzzy control to a batch type reactor will be described.

The mainstream temperature control for batch reactors is a cascade from the reactor internal temperature to the jacket temperature using a PTD controller. In particular, in reaction temperature control, since a wide variety of reactants and catalysts are involved, there are many methods in which experientially obtained knowledge is combined with sequence operation and PID control.

Here, it can be easily inferred that fuzzy control is effective from the effectiveness of empirical rules, but in the current literature, for example, rThe
Application of Fuzzy Cont
r.

5ysten+ to Industrial Pro
cess J 1976 Maldani et al.
I can't find any chemical-related applications.

Referring to FIG. 10, an example of a batch reactor control system using a reactor control unit 1 having a fuzzy controller 101 having a membership function according to the conventional method described above will be considered.

2 is a batch reactor, 3 is a heating jacket surrounding this, 4 is a heating steam S supply pipe line to the jacket,
5 is a pipe for supplying cold water W for cooling to the jacket, 6 is a steam flow control valve, 7 is a flow control valve for cold water, 8 is a discharge pipe from the reaction vessel, and 9 is a supply pipe for catalyst M. .

10 is a pressure sensor for the pressure P in the can, 11 is a temperature sensor for the temperature T of the reaction liquid in the can, and 12 is a temperature sensor for the temperature T in the jugget. The measurement signals P, T, , T of these sensors are led to a control unit 1 .

In the control unit 1, 102 is the current value T of the measured value T.
I(n) and the measured values of past multiple sample periods are T1(n
-1), Tlfn-2)--... is a time series data holder.

103 is the current value 'r'2(n) of the measured value TJ and the measured values of the past multiple sample periods T2(n-1), T2(n
-2) It is a time series data holder that holds...

104 indicates the current value p (n) of the measured value P and the measured values of the past multiple sample periods as P (n-1) and P (n-2
)... is a time series data holder that holds...

105 is the measured value and the set value SV given programmatically.
The current value of the deviation E (n) -T1 (n) -3V and the deviation of the past multiple sample periods are E (n-1) and E
It is a time series data holder that holds (n-2)...

106 is the rate of change of deviation E (amount of change in one sample period) Δ
E(n) -F, fn) -E(n-1) or rate of change of measured value (change amount of j sample period) ΔE (n)T
It is a time-series data holder that holds deviations of past plural sample periods of (n)-T(n-1) as ΔE(n-1), 8E(n-2), and so on.

107 is the automatic or manual operation output M of the controller 101
Current value of V MV (n) and operation outputs of past multiple sample cycles MV[n-1), MV[n-2) =-・
It is a time series data holder that holds .

Based on the time series data of these data holders, the fuzzy controller 101 determines the optimum parameters of the membership function, and transmits the operation output MV to the control valves 6 and 7 by fuzzy calculation based on the membership function.

<Problems to be Solved by the Invention> The difficulty in controlling the temperature of a reaction vessel lies in the fact that endothermic reactions and exothermic reactions are included in the process, and the timing of their occurrence is difficult to grasp.

Expressing this in control terms, it means that dead time (L), lead time (T'), and process gain (G) change.

Now, suppose that a change in process characteristics can be identified by some method. In the fuzzy controller, the parameter (■5
．． The shape of the control rule and membership function (NB, NS, ZE, PS, PB) corresponding to the size of T, G), each input/output variable (deviation E, rate of change ΔE, increment ΔU of manipulated output) You have to readjust everything every time.

This type of adjustment work becomes extremely complicated in reactors and the like where process characteristics frequently change over time.

In conventional fuzzy control, the above-mentioned known document "Fuzzy Control J P1" by Michio Kanno, published by Nikkan Kogyo Shimbun, 1988,
13 to P122, the idea of fuzzy adaptive control is introduced.

In this method, for example, the membership functions at dead times Ll and L2 are adjusted in advance, and when the dead time is identified as L (L1≦L≦1-2) due to a characteristic change, the corresponding membership functions are adjusted in advance. Interpolation ratio (1, -Ll) +
(LL2). T,G
When , the membership function is corrected by interpolation in the same way.

This type of interpolation method is formal, and it is unreasonable to deal with changes in characteristics only by adjusting the membership function.
Furthermore, as the number of parameters increases, the number of membership functions that must be adjusted and set in advance also increases, so the configuration of the data holder, etc. shown in Figure 10 becomes complicated, and the adjustment work becomes extremely complicated. .

The present invention aims to provide a fuzzy control method that solves these problems.

Means for Solving the Problems〉 The feature of the method of the present invention is that the operation output obtained by performing fuzzy calculations on the deviation between the measured value and the target value of the process at a predetermined sampling period is applied to the process at a predetermined sampling period. In a fuzzy controller with a fixed control rule and membership function as well as supplying the above process characteristics (time, lead time, process gain)
is identified, and the sampling period of the fuzzy controller is changed based on this identification result.

A second feature is that the characteristics of the process (delay time, lead time, process gain) are identified, and the sampling period and gain of the fuzzy controller are changed based on the identification results.

(Operation) The fuzzy controller supplies the process at a predetermined sampling period with a manipulated output obtained by performing a fuzzy calculation on the deviation between the measured value and the target value of the process at a predetermined sampling period.

The control rules and membership functions in the controller are fixed, and the sampling period of the fuzzy controller is changed depending on the characteristics of the identified process (time, lead time, process gain).

Both the sampling period and the gain are changed as necessary.

(Example) An example of a fuzzy control device implementing the method of the present invention will be described based on FIG.

In the control unit 1, 108 is a fuzzy controller into which the deviation E is input, and 109 is a Δt for changing the control sampling period Δt and gain K of the controller.
, is a change function.

Reference numeral 110 denotes a process characteristic (L, T, G) estimation means, which drives the change function 109 to Δt based on the identification result.

Various methods have been proposed for identifying the dynamic characteristics of a process, and the present invention assumes that the dynamic characteristics can be estimated using these methods.

The control rule for the deviation and its rate of change in the fuzzy controller 108 is fixed, and the membership function is also fixed.

The function of the change function 109 to Δt is as follows: (1) The sampling period Δt is increased in response to an increase in dead time.

(2) Decrease the controller gain K in response to an increase in the process gain G.

②In response to an increase in lead time T, adjust the increase or decrease by combining Δt and K.

The control rule that realizes these functions is changed to Δt, Function 1
It will be added in 09.

Therefore, the control rules are nine well-known fixed rules regarding the deviation E and its rate of change ΔE in the fuzzy controller: (1) IF E=NB &ΔE=ZE, THN Δ
U=NB(2) IF E=NS &ΔE=ZE, TH
In addition to N ΔKo N5 (9)..., the following four control rules are added to the change function 109 to Δt.

(10) IF L=SMALL & G=SH^11
．． THEN Δt=sH^1-L&=BIG (11) IF L=SHAIL & G=b+a, T
HEN Δt=sHALL & ni=SHALL (12) IF L=BIG & G=S14ALL,
THEN Δt=BIG &ni=BIG (13) IF L=BIG 8 G=BIG, THE
N Δt=BIG &=SHALL The criteria for determining 514ALL and BIG for each of the above parameters are determined based on the operator's experience and experiments.

If the lead time changes, an additional rule is created by further combining Δt and K. In this case, glue, T, and G SH^11. SMALL of Δt and K corresponding to BIG
, BIG also has operator experience9. Determined based on experiment.

The above explanation is based on the identification results of process characteristics Δt
Although a case has been described in which both of and K are changed, changing only the sampling period Δt is also effective in controlling the reactor, and the configuration can be simplified.

FIG. 2 shows an embodiment in that case, and the same elements as those explained in FIG. 10 are denoted by the same reference numerals, and the explanation will be omitted, and the characteristic parts will be explained. The process characteristic estimating means 110 inputs the pressure measurement value P, the temperature measurement value T, the catalyst M, other time information, brand information, etc., identifies the process characteristic, and changes Δ of the fuzzy controller 109'. drive,
The Δ changing function 109' changes the sampling period Δt according to the set control rule.

Next, a configuration example of the characteristic estimating means 110 in the reactor process will be described.

As shown in Figure 3, the estimation mechanism is as follows: time 1 reaction heat ΔH 1 reaction liquid temperature T1
．． Jaguette temperature Tj, various parameters? The reaction rate XA is calculated by inputting the meter, and based on this result, the process gain G is estimated, and the dead time and lead time T are estimated.

First, the model equations for the reaction vessel are expressed by the following equations (1) and (2).

d　x　A/　d　t (0A10AO)k.

xEXP (-E/RT, ) (1) (1/CP
ρ ) ・ −Δ H−C＾ko −EXP
(-E/RT,) 10U-A (T, -T, > (
2) dT, /dt: Initial concentration of substance A (mol/m''): Concentration of substance A (mol/m) Two-speed variable constant: Activation energy (cal/mol): Gas constant (cal/mol・K) CP: Specific heat of reaction liquid (cal/kg-K)
ρ: Reaction liquid density (kg/m3) ΔT: Reaction heat (- sign indicates endotherm) T: Jacket temperature (6K) T, : Reaction liquid temperature (°K) U: Heat transfer coefficient (cal/ m3-s-K)A:
Heat transfer area (m2)

Furthermore, X A , T J ; T i , ΔH are functions of time t, and x^ (t), 'rJ (tl , 'r'
,(t).

ΔH(t).

FIG. 4 shows the change characteristics of the reaction rate XA with respect to time t as a result of solving the above equations (1) and (2).

Figure 5 is a table for estimating the process gain G from the relationship between the reaction rate XA and its rate of change dx^/dt determined from the characteristics in Figure 4, and the current process gain is identified by referring to this table. .

When there is dead time in the process, the third term on the right side of equation (2), U-A (T, -T,), is changed to U-A (T
j (t-L)-Ti (t-L)) and consider L as one of the model parameters. Based on the data obtained from time to time, dead time is identified in real time using table means in the same manner as in the case of gain estimation.

The lead time T can also be identified by a similar procedure.

Another method for identifying process characteristics will be explained with reference to FIG.

FIG. 6 is a characteristic diagram showing the temperature control results of the reaction vessel controlled by normal PID control, and (a) is a temperature rise curve showing the relationship between the reaction liquid temperature T and the program setting value Sv.

(b) is a characteristic diagram showing the relationship between jacket temperature T and operation output MV at batch start time t. M from to tl
A full-open V sequence is executed, and from t to t2, M
Constant value control of V preset is executed, and automatic control is performed after t2.

(C) is a characteristic curve showing the change in the amount of heat generated in the can, (d)
is a characteristic curve showing the pressure change inside the can.

Based on the actual example of such a characteristic curve, the change rate of process gain and (a), (b) are calculated from the actual curve of MV value in (b).
Estimate changes in dead time and lead time over time from the relationship.

<Effects of the Invention> (1) The conventional method of identifying the characteristics every time the process characteristics change and adjusting the parameters of the membership function of the fuzzy controller to the optimal value based on the results is extremely complicated. becomes.

As in the present invention, the fixed value control rule and membership function used before changing the process characteristics are fixed as they are, and instead, the sampling period or sampling period and controller gain are made variable, thereby simplifying adjustment. At the same time, it becomes possible to improve the controllability of adaptive control.

(2) When creating membership functions and control fuzzy control rules, it is possible to think of static behavior (if things continue as they are, this is where the future will settle down), and to avoid temporal relationships that humans are not good at. System design becomes easier because rules can be determined without understanding changes in characteristics, that is, without describing dynamic characteristics.

(3) It is a common principle among experienced operators to control fast response processes with short sampling times and slow response processes with long sampling times, and by making this independent from the fuzzy inference section, inference The software configuration of the parts can be made smaller and simpler.

[Brief explanation of the drawing]

Fig. 1 is a basic block diagram of a fuzzy control device to which the method of the present invention is applied, Fig. 2 is a block diagram showing an embodiment of a reactor control device to which the method of the present invention is applied, and Fig. 3 is a diagram of a reactor process model.
Figure 4 is a characteristic diagram showing the change in reaction rate over time, which shows the result of solving the model equation, Figure 5 is a table configuration diagram that estimates process gain from a combination of reaction rate and its rate of change. , Fig. 6 is a characteristic diagram showing the PID temperature control results of the reactor, Figs. 7 to 9 are explanatory diagrams regarding membership functions and fuzzy operations in fuzzy control, and Fig. 10 is a diagram showing membership functions according to the conventional method. Reactor control unit 1 having a fuzzy controller 101 with
Fig. 1 is a configuration example of a batch reactor control system according to Fig. 1. 1... Control unit l~, 108... Fuzzy controller 1
～Roller, 109... Δt, change function, 110...
・Prose →] Ward i group Ward group

Claims (2)

[Claims] (1) A fuzzy system in which a fuzzy operation is performed on the deviation between the measured value and the target value of the process at a predetermined sampling period, and the operation output is supplied to the process at a predetermined sampling period, and the control rules and membership functions are fixed. A fuzzy control method, characterized in that the characteristics of the process (dead time, lead time, process gain) are identified in the controller, and the sampling period of the fuzzy controller is changed based on the identification result. (2) A fuzzy system with a fixed control rule and membership function, in which the manipulated output obtained by performing fuzzy calculations at a predetermined sampling period on the deviation between the measured value and target value of the process is supplied to the process at a predetermined sampling period, and the control rule and membership function are fixed. A fuzzy control method, characterized in that the characteristics of the process (dead time, lead time, process gain) are identified in the controller, and the sampling period and gain of the fuzzy controller are changed based on the identification results.