JPH03100704A - Process controller - Google Patents

Abstract

PURPOSE:To find a control constant in a short time without generating any disturbance to a process by finding the outputs of a modeling part and an estimation accuracy arithmetic part or the control constant of a control arithmetic part. CONSTITUTION:A data gathering part 22 holds a manipulated variable and a process quantity as time-series data for a constant time and a calculation command part 26 actuates the modeling part 23 when the process quantity and manipulated variable vary above a specific value. The internal parameter adjustment part 233 of the modeling part 23 adjusts the internal parameter of a process model 232 so that the output of a process 21 matches the output of a process model 232 and an estimation accuracy arithmetic part 24 calculates the accuracy of the process model 232. A control constant arithmetic part 25 calculates the control constant from those outputs and outputs it to a control arithmetic part 20. Consequently, the control constant is set again without generating any disturbance to the process.

Description

DETAILED DESCRIPTION OF THE INVENTION <Industrial Application Field> The present invention relates to a process control device for controlling a process, and particularly to a process control device having a function of automatically adjusting a control constant to the maximum IN value.

<Prior Art> In order to control a process to be controlled under optimal conditions, it is necessary to obtain the dynamic characteristics of this process and sequentially obtain control constants for calculating the manipulated variable.

As such process control devices, open loop type, limit sensitivity type, response waveform type, and adaptive control type process control devices are known.

FIG. 12 shows the configuration of an open-loop process control device.1 In normal control, the switch 1 is placed on the control calculation section 2 side. The set amount and the process amount are inputted to the deviation calculating section 3 to calculate the deviation between them, and the control calculating section 2 calculates the manipulated variable by PID calculation or the like and outputs it to the process 4. At regular intervals or when the deviation becomes large, the switch 1 is set to the identification signal generator 5 side, and a step signal or a pulse signal is applied to the process 4. From the response of the process 4 at this time, control constants are calculated by the 7ieglar-Nichols method or the like and set in the control calculation unit 2.

FIG. 13 shows the configuration of a limit sensitivity type process control device. The output of the control calculation section 2 and the output of the limit cycle generation section 6 are added in an addition section 7 and outputted to the process 4 as a manipulated variable. The limit cycle generating unit 6 periodically changes the manipulated variable input to the process 4 between a maximum value and a minimum value,
From the limit period and limit sensitivity at that time, Ziegler-N
Control constants are calculated using the ichhols method or the like and set in the control calculation section 2.

FIG. 14 shows the configuration of a response waveform type process control device, in which set amounts and deviations are input to response waveform rule 8. The response waveform rule 8 is composed of tuning rules such as Fuzzy Ru1e, and this rule is applied to the overshoot of the process amount, the damping ratio, the period, etc. to obtain the control constant and set it in the control calculation unit 2. .

The configuration of an adaptive control type process control device is shown in FIG. 15. The output of the control calculation unit 2 and the output of the identification signal generation unit 9 are added by an addition unit 10 and input to the process 4. Also,
The process amount and the output of the addition section 10 are input to the dynamic characteristic identification section 11, and the control constants are estimated and output to the control calculation section 2. That is, a slight disturbance is generated in the identification signal generation section 9, and a control constant is determined from the dynamic characteristics of the process amount at that time and set in the control calculation section 2.

<Problems to be Solved by the Invention> However, such process control devices have the following problems. In open-loop and adaptive control types, an identification signal must be given to the process, and in limit-sensitivity types, sustained vibrations must be generated, that is, disturbances must be applied to the process.
Restrictions occur when trying to determine control constants during operation. In addition, the response waveform type requires waveform observation several times in order for the control constant to converge, and it takes a certain amount of time.

Therefore, neither control device could freely reset the control constants.

<Object of the Invention> An object of the invention is to provide a process control device that can reset control constants in a short time and without causing any disturbance to the process.

Means for Solving the Problems> In order to solve the above problems, the present invention collects the manipulated variables and process variables to be output to the process at a predetermined time by a data collection unit, and collects the manipulated variables and process variables to be outputted to the process, and then adjusts the data collection unit to match the collected process variables. The modeling unit adjusts the internal parameters of the process model, and the estimation accuracy calculation unit calculates the accuracy of the process model, and the modeling unit is activated when the manipulated variable or process variable fluctuates by a predetermined value or more. The control constant of the control calculation section is determined from the outputs of the modeling section and the estimated accuracy calculation section.

Further, the gain constant of the process model is determined from the output of the process model and the process amount.

In addition, the simplex method is used for the search, and the length of the side components of the initial polygon is made longer than the sampling period collected by the data collection unit, and after the first convergence, the simplex is recreated at that point. This is what I did.

Furthermore, a stepwise manipulated variable is applied to the process at the beginning of control, and the sampling period and initial values of the control constant are determined from the response of the process variable at that time.

Control constants can be determined in a short time without causing any disturbance.

<Embodiment> FIG. 1 shows an embodiment of the process control device according to the present invention. In FIG. are input, and the manipulated variable is calculated by performing proportional-integral-differential calculations on these deviations.

21 is a process to be controlled, and the control calculation unit 20
The manipulated variable, which is the output of , is input, and the process variable is output. Reference numeral 22 denotes a data collection unit into which the operation amount and process amount are input, and holds these data as time-series data for a certain period of time. 23 is a modeling section, which includes a preprocessing section 231, a process model 232, an internal parameter adjustment section 233, and a comparison section 234. The output of the data collection unit 22 is subjected to preprocessing such as filtering in a preprocessing unit 231, and the manipulated variable is processed by the process model 23.
2, the process amount is input to the comparison section 234. Process model 232 is a simulation of process 21. The output of the process model 232 is input to a comparing section 234, where it is compared with a separately input process quantity.

This comparison result is input to the internal parameter adjustment section 233. The internal parameter adjustment unit 233 adjusts the internal parameters of the process model 232 so that the output of the process 21 and the output of the process model 232 match. Reference numeral 24 denotes an estimation accuracy calculation unit, into which the output of the modeling unit 23 is input, and calculates the accuracy of the process model 232. 25
is a control constant calculation section, and the internal parameter adjustment section 233
and the outputs of the estimated accuracy calculation unit 24 are input, and control constants are calculated from these outputs and output to the control calculation unit 20.

The control calculation section 20 calculates the manipulated variable based on this new control constant. Reference numeral 26 denotes a calculation command section, into which the process amount and operation amount are input, and when these change by a predetermined value or more, by operating switches 27 to 30, a calculation command is output to the modeling section 23, and the modeling section Start the operation of 23.

Next, the operation of this embodiment will be explained based on the flowchart of FIG. Note that, irrespective of this operation, the data collection unit 22 collects the operation amount and the process amount as time-series data at regular intervals.

In FIG. 2, the calculation command unit 26 reads the process amount (pv) and the manipulated variable (MV) at regular intervals, and determines whether the fluctuations thereof are larger than a predetermined value. If the fluctuation is small, the calculation ends without issuing a calculation command. This judgment is made by calculating the area (shaded area) by integrating the deviations of the process variables and manipulated variables from the steady state, as shown in FIG. The time-series data (PV) and (MV) of the process amount and operation amount collected by the unit 22 are read, and filtered by the preprocessing unit 231 to remove stationary components and noise. This filtering is performed, for example, by the following calculation, assuming that the time series data of the process amount is PV4, and the steady portion is Pvo.

PV, = (1-α) ・PV, +a-(PV-P
Vo) α: Filter constant Filtered time-series data Pν, Hv[ of process variables and manipulated variables are input to the process model 232, and the model output is calculated. This calculation differs depending on the process model, but if the nth model output is turned on, for example, 0 = β8・0o−1 + (1−β14)・KM・HVF(n−LH)・・
(1) KH: Process model gain LH: Process model dead time βN = Process model first-order lag coefficient HVF (n-Ls): Value of MV before LM time It is done by calculation. This operation reaches the predetermined maximum number of repetitions, or the output On of the process model 232 becomes the time series data Pv of the process amount.
This is done until the value approaches 1. If the number of repetitions does not reach the maximum number of repetitions, and as shown in FIG.
If the difference with F is not small, make sure that this difference becomes small (
That is, the gain KN and dead time LH1 - order delay coefficient β8 of the process model 232 are adjusted so as to move in the direction of the arrow in FIG. 4, and the process model 232 is started again. The maximum number of repetitions is reached or process model 2
When the difference between the output On of 32 and the time series data Pv of the process amount becomes sufficiently small, the estimated accuracy calculating section 24 calculates the accuracy K of the process model 232. This accuracy is determined by, for example, the following equation (2) or (3).

K= 1− [Σ(PV, (i) −0(i) )
22=Σ(PV(i))/Σ(0(i)
) −(3) However, PVr h) , 0 (
i) Time-series data of the process amount at each time point i and the output of the process model 232. If this estimation accuracy is higher than a predetermined value, the control constant calculation unit 25 calculates proportional, integral, and differential constants, that is, control constants, from the obtained process model 232 of the first-order lag model using the 2iegler-Nichols method. The control calculation unit 20 uses this newly determined control constant to calculate the manipulated variable. Note that the process model does not necessarily have to include a first-order lag system as shown in equation (1) above, and if there is a measurable disturbance such as load fluctuation in the process 21, this disturbance can be used in the model calculation. May be used for

In the embodiment shown in FIG. 1, the internal parameters of the process model 232, that is, the gain KM, the dead time LH1 - the order lag coefficient βH, all have to be repeatedly calculated by changing their values little by little. The problem was that it took a lot of time. An embodiment that solves this problem will be described below.

In this embodiment, the gain is obtained from the integration ratio of the process amount and the output of the process model when the process gain is set to 1, and the dead time Ls and the -order delay coefficient β8 are obtained using an iterative search method. It is something. FIG. 5 shows the configuration of this embodiment. However, since the parts other than the data collection section 22 and the modeling section are the same as in FIG. 1, their description will be omitted.

In FIG. 5, 31 is a modeling section, which has the same function as the modeling section 23 in FIG.

311 is a preprocessing section, which functions in the same way as the preprocessing section 231 in FIG. That is, the process amount and operation amount data held in the data collection unit 22 are filtered to remove stationary noise. 312 is a process model,
Similar to the process model 232, an actual process is simulated. The manipulated variable preprocessed by the preprocessing unit 311 is input to this process model 312 . 313 is a gain calculation unit, into which the output and process amount of the process model 312 are input, and calculates the gain of the process model 312.

314 is a comparison unit to which the output of the process model 312 and the process amount are input, and these values are compared. 31
Reference numeral 5 denotes an internal parameter adjustment unit, into which the output of the comparison unit 314 and the output of the gain calculation unit 313 are input, and changes the internal parameters of the process model 312 so that the difference between the output of the process model 312 and the process amount becomes sufficiently small. Make it. The output of the internal parameter adjustment section 315 is input to the estimated accuracy calculation section 24 and the control constant calculation section 25 shown in FIG. 1, and the control constant of the control calculation section 20 is determined. The calculation command unit 26 controls the switches 32-35.

Next, the operation of this embodiment will be explained based on FIG.

First, the calculation command unit 26 shown in FIG. 1 periodically reads the manipulated variables and process variables, and performs the following processing only when the fluctuations in them are large.

That is, after the preprocessing unit 311 filters the time series data (PV) and (MV) of the manipulated variables and process variables to remove stationary noise, the process model 312 calculates the output of the process model. In this calculation, the process gain is set to 1, and the following equation (4) is used.

0 = β8・0o-1 + (1-β) ・HVF < n LM) ・=
(4) However, the meaning of each coefficient is the same as in equation (1) above.

Next, the gain calculation unit 313 calculates 8 as the process gain using the following equation (5).

K = [Σ(0(n) ・PV, (n) ) /
<Σ0(n) )・・・・・・・・・ (5) However, O(n) is the value obtained by the above equation (4), that is, the output of the process model 312 when the process gain KH is set to 1. It is. This calculation reaches a predetermined maximum number of repetitions, or the product of the output On of the process model 312 and the process gain KH calculated by the gain calculation unit 313 becomes the process amount time series data Pv,
This is done until it is close enough to . If none of these conditions are met, adjust the dead time L and -order lag coefficient β8 of the process model 312 so that this difference becomes small, start the process model 312 again, and output according to the equation (4) above. is calculated, and the process gain KN is calculated by equation (5).
When the zero condition of repeating the operation of calculating K is satisfied, the estimated accuracy calculation unit 24 calculates the accuracy K of the process model 312
is calculated, and if this estimation accuracy is smaller than a predetermined value, the control constant calculation unit 25 calculates Z ieg 1er from the process model 312 of the obtained first-order lag model.
-N 1cho Is method to calculate proportional, integral, and differential constants,
That is, control constants are calculated. The control calculation unit 20 uses this newly determined control constant to calculate the manipulated variable.

Note that the internal parameter adjustment unit 315 calculates the dead time L8 and the -order lag coefficient β8 by an iterative search method using the simplex method. Furthermore, if the process gain KH is required at the beginning of repetition, the previous process gain KH value is used. In this embodiment, the process gain K is automatically determined when the dead time Lr1 and the -order lag coefficient βI are determined, so that the number of repeated searches can be reduced.

As a method for the internal parameter 11 integer to search for the internal parameter of the process gain, for example, the simplex method can be used. The simplex method takes several geometrically arranged points on R and searches by comparing the values of the objective function at those points.
For example, it is described on pages 284 to 287 of Nonlinear Programming (authored by Hiroshi Konno and Hiroshi Yamashita) published by Nikkei.

However, in the normal simplex method, since the input and output data are in principle discrete values, the objective function also becomes a space of discrete values, and there is a problem that the search cannot be performed if the search point enters the subspace. There was also a constraint that parameters such as time constant and dead time must be positive numbers. 7th
The figure shows the search procedure using the simplex method and eliminating the constraints described above. In this figure, the process data collected by the data collection unit 22 is defined as a geometric point X on R, and this Let the point function f□ be the objective function. The search is performed by comparing the values of the objective numbers f<>.
Also, the point that gives the maximum value of f() among the points Furthermore, 1111 images x −<
10(2) ・x” −axhα〉0 Expansion 'J x'=y−x'+(1−r)−x”γ〉1 Contraction βε(0,1
) Reduction Reduce all vertices in the xl direction.

Define. In FIG. 7, after initialization, the first simplex, that is, the 0th order that generates (n+1) affine independent white colors on R, xlxS, x
', x' are determined. Then, it is determined whether it converges. Judgment of convergence is 'f'(f (x') - f'')2
/(n+1)≦ε4゛龜l is satisfied, however, f11=Σf (x' )/(n+1)8I When the convergence condition is satisfied, if it is the first convergence, then Generate the first simplex at the point xlxS
, x', x'', repeat M times. If convergence is achieved for the second time or more, the process ends. If it does not converge, x',
Find f(x) and judge whether f(x')≦f(xS) is satisfied. If it is not satisfied, f (x
' ) < f (xh), and if this is satisfied, xh is xr
If it is not satisfied, leave it as is and replace it with Xo and f(x
c) and determine f (x)<f (xh). If this formula is satisfied, xh is replaced by Xo, and if not, xl is (xi + x ) /
Change ttk at 2 and return to the determination of xlxS, x' and xl. On the other hand, if f(x)≦f(xS) is satisfied, it is determined whether f(x)≧f(x') is satisfied, and if not, x'
Find f(x'), judge whether f(X) <f(X') is satisfied, and if satisfied, replace xh with X; if not, f(x) ≥ f(x') When is satisfied, xh is replaced by X and x, xS
, x', returning to the determination of x. In this example, compared to the normal simplex method,
By generating a simplex at that point during the first convergence and applying the simplex method again, it is possible to prevent entry into the subspace and also to apply negative parameters. Note that in this embodiment, the length of the side components of the initial polygon is made longer than the sampling period, and the initial polygon is recreated at least once.
You may perform only one of them.

Fig. 8 shows yet another embodiment. In the embodiment shown in Fig. 1, the initial values and sampling period of the proportional, integral, and differential coefficients must be set prior to control operation. If these settings are not appropriate, the control Automatic adjustment of constants is not successful, resulting in poor control characteristics.

In this embodiment, in order to avoid such a situation, the control constants and sampling period are tuned in advance. Note that the same elements as in FIG. 1 are given the same reference numerals and their explanations will be omitted.

In FIG. 8, 40 is a step input applying means;
A stepwise operation amount is applied to the process 21. By inputting this manipulated variable, the process amount gradually increases as shown in FIG. Reference numeral 41 denotes a first monitoring means, which monitors the process amount which is the output of the process 21, and after the stepwise operation amount is applied by the step manual force application means 40, the process amount is set to a first predetermined value, for example, full scale. 1% of
The time width t, that is, the first time width, is measured. Reference numeral 42 denotes a sampling period calculation means, into which the first time width measured by the first monitoring means 41 is input, and from this value the sampling period of the data collection section 22 is calculated. 43 is a second monitoring means, which monitors the process amount which is the output of the process 21, and after the step input application means 40 applies a stepwise operation amount, the process amount is larger than the first predetermined value. The time width t2, ie, the second time width, until the second predetermined value, for example, 2% of the full scale, is reached is measured. Reference numeral 44 denotes a sampling period correction means, into which the second time width measured by the second monitoring means 43 is input, and the sampling period calculated by the sampling period calculation means 42 is corrected.

Reference numeral 45 denotes a switch constituting the sampler, which is initially turned on for each period calculated by the sampling period calculating means 42 and later for each period corrected by the sampling period correcting means 44, so that the process amount data is transferred to the data collecting section 22. Collected by. This collected data is input to the modeling section 23, which adjusts its internal parameters so that the output of the process model matches the collected process amount as explained in FIG. is calculated.

Next, the operation of this embodiment will be explained based on the flowchart of FIG. This operation is executed at regular intervals in the initial stage of control. First, it is determined whether it is an initial start, and if it is an initial start, a step-like operation amount is applied to the process 21.

If it is not the first step, monitor the process volume and
Determine whether any of the following three steps apply. If none of the steps apply, the process ends without doing anything.

■When the process amount PV exceeds 1% of full scale.

■When the process amount PV exceeds 2% of full scale.

(2) When the process amount Pv settles to a constant value or becomes an integral multiple of the sampling period determined by the sampling period correction means 44.

If step (■) applies, measure the elapsed time t1, that is, the first time width, after applying the step-like operation amount to the process, and set one integer fraction of this time as the sample time t0. For example, t = t, /4, and if step ■ corresponds to
It is determined whether this time width t2 is larger than twice the time t1 obtained in step 1. If it is larger than twice tl, t = t2/4 is set, and the already determined process quantity is interpolated to correct the process quantity to the sampled value at this new sampling time t. If step (2) is applicable, the process quantity data collection is finished, the time series data (p, v), (MV) of the process quantity and operation quantity accumulated in the data collection unit 22 are organized, and this time series data is Based on modeling part 23
By adjusting the internal parameters of the process model 232, the control constants are determined. This control constant is set as the initial value.

From then on, data will be collected at the cycle of this sample time t3,
The control constants are calculated according to the present invention explained in FIG.

Note that in this embodiment, a step-like operation amount is applied only once at the beginning of control, but in the case of a stereotaxic system, the first
Although the process amount PV converges as shown in FIG. 1(A), in the case of an integral system, an offset remains as shown in FIG. 1(b). Therefore, after the process amount has stabilized, a step signal is applied in the opposite direction to remove the offset, and then sampling of process data is continued. Furthermore, if the change in the process amount exceeds the allowable range, a step signal is applied in the opposite direction even before it is stabilized to prevent the process amount from becoming excessive. Furthermore, it can also be determined from the offset amount whether the process is an integral system or not. In FIG. 11(b), the time 'r for applying the stepwise manipulated variable is set such that the predetermined time or the amount of change in the process amount is greater than ΔPImax.

Further, in these embodiments, the control constants are obtained from the internal parameters of the modeling section by the Z ieg far, -N 1cho Is method, but the least squares method or the maximum likelihood method may also be used.

<Effects of the Invention> As explained above in detail based on the embodiments, in this invention, process data is accumulated in the data collection section, and only when a change occurs in the process data, the modeling section is activated to update the process data. The internal parameters of the model were adjusted, the estimated accuracy of the process model was calculated, and the control constants were determined from these. Therefore, since the identification signal is not required, there is an effect that no unnecessary disturbance is caused to the process.

Furthermore, since the estimation accuracy of the process model is calculated, when it is determined that the model is inappropriate, adjustment of the control constants can be reduced or stopped, thereby preventing the process control device from running out of control.

Further, since the process model is calculated while the process data is temporarily accumulated in the data collection unit, time for data collection is not required.

Further, by calculating the process gain from the process amount and the output of the process model, the search time can be shortened.

Furthermore, since the simplex method is used as the search method and the simplex is re-created by f& of the initial convergence, the drawback of entering a subspace can be eliminated.

Furthermore, since a stepwise manipulated variable is given to the process at the beginning of control, and the initial values of the sampling period and control constant are determined from changes in the process variable at that time, more appropriate control can be achieved.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the process control device according to the present invention, FIG. 2 is a flowchart for explaining its operation, and FIGS. 3 and 4 are for explaining the operation of the embodiment shown in FIG. FIG. 5 and FIG. 6 are diagrams showing the configuration and operation of other embodiments, FIG. 7 is a flowchart of a search method using the simplex method, and FIGS. 8 to 11 are diagrams showing other embodiments. FIGS. 12 to 15, which illustrate embodiments, are configuration diagrams of a conventional process control device. 20... Control calculation unit, 21... Process, 22...
・Data collection department, 23...Modeling department, 24...
Estimated accuracy calculation unit, 25... Control constant calculation unit, 26...
- Calculation command center, 40... step input application means, 41
... first monitoring means, 42 ... sample period calculation means, 43 ... second monitoring means, 44 ... sample period correction means, 232.312 ... process model, 2
33,315...Internal parameter adjustment section, Fig. 1 Fig. 3 Back Fig. 5 Fig. 8 Fig. 9 Fig. 1/Fig. 72 Fig. 74 Fig. 75

Claims (5)

[Claims] (1) A control calculation unit into which process variables and set values are input, which calculates a manipulated variable of the process from these values and outputs it to the process; and a data collection unit which collects the manipulated variable and process variable for a predetermined period of time. and a process model therein, into which the manipulated variable collected by the data collection section is input, and whose internal parameters are adjusted so that the output of the process model most closely matches the process variable collected by the data collection section. It has a modeling section that performs adjustment, and an estimated accuracy calculation section that calculates the accuracy of this process model, and starts the modeling section when the manipulated variable or process amount fluctuates by a predetermined value or more, and operates the modeling section and A process control device characterized in that a control constant of the control calculation section is determined from an output of the estimated accuracy calculation section. (2) A gain constant, which is one of the internal parameters of the process model, is determined from the output of the process model and the process amount when the gain constant of the process model is set to 1. Process control equipment as described in Section. (3) A simplex method is used to adjust the internal parameters of the process model, and the length of the side components of the initial polygon is made longer than the sampling period collected by the data collection unit. The process control device according to item 1. (4) A claim characterized in that the simplex method is used to adjust the internal parameters of the process model, and after the internal parameters converge, the initial polygon is re-created and searched at least once using the coordinates. The process control device according to item 1. (5) A control calculation unit into which process variables and set values are input, which calculates a manipulated variable of the process from these values and outputs it to the process; and a data collection unit which collects the manipulated variable and process variable for a predetermined period of time. a step input applying means for applying a step input to the process; a first monitoring means for monitoring a first time width until the process amount changes by a predetermined first level; and a step input applying means for applying a step input to the process; a sampling period calculating means that uses a value obtained by dividing the width as a sampling period, and samples the process amount by the data collection unit in this sampling period; a second monitoring means for monitoring the second time width; a sampling period correction means for correcting the sampling period calculated by the sampling period calculation means from the second time width and the first time width; and a modeling section that calculates control constants using a process model from the process data collected by the section, and at the beginning of control, the step input application means applies a step-like manipulated variable to the process, and the time-series data of the process at that time is applied. A process control device characterized in that a modeling section is activated to obtain control constants.