JPS624724B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a PI or PID controller that automatically determines each parameter on operating characteristics, that is, a transfer function, to an optimal value in accordance with a controlled object.

For example, a multi-capacity process such as the temperature control system of an electric furnace or a multi-capacity process with dead time (1
As is well known, a PI (proportional + integral action) regulator or a PID (proportional + integral + differential action) regulator is often used in a feedback control system that controls a (secondary delay system).

When incorporating a general-purpose PI or PID controller into a control system for a specific controlled object, the parameters K _P , T _i , T _d (hereinafter referred to as PID constants) on the operating characteristics of the controller are set to the control system of the controlled object. It must be adjusted to the optimum value according to the process characteristics. Conventional regulators are as above.
Only knobs and the like are provided to vary the PID constant, and the optimum adjustment described above requires separate measurement of the characteristics of the controlled object and manual adjustment and setting accordingly. In other words, according to the limit sensitivity method, the characteristics of the system are measured by applying an appropriate amount of operation to the controlled object, and based on the measured values, the appropriate value of the above PID constant is calculated, and it is set in the controller. .
In the past, extremely troublesome operations were required for optimum adjustment, and although there would not be much of a problem if the regulator was incorporated into a certain control system and used for a long period of time, If changes are made frequently, or if the characteristics of the system change due to changes in the surrounding conditions (temperature, etc.) even if the control target is the same, the characteristics of the system must be measured and readjusted each time. However, there were many practical problems.

In view of this problem, the present inventors first applied a constant amount of operation to the controlled object in a step manner with the regulator incorporated in the controlled object, measured the step response, and calculated the dead time and Calculate the time constant slope,
We have developed a controller with a function to automatically set the PID constant determined based on these two values, and have already applied for a patent.

When the control target is a multi-capacity localization type process with dead time, the dead time and time constant slope on the step response characteristics described above are values that very well represent the characteristics of the process, and the PID constant can be calculated based on these values. By determining this, considerably highly accurate control can be expected.

However, many control objects in reality are not processes to which the theoretical equations perfectly apply, but instead include various nonlinear elements. Therefore, in the case of the previously proposed regulators mentioned above, even if optimal control can be performed for a certain target value using the determined PID constant, control accuracy may decrease when the target value changes. often seen. The reason for this is that although the responsiveness of the controlled object near the target value changes depending on the target value, the PID constant is determined based on the dead time and time constant slope of the step response, ignoring the target value. It lies in being.

The present invention has been made based on the above considerations, and provides a regulator in which an optimal PID constant is determined according to a target value.

That is, the regulator of the present invention includes step response measuring means that increases the amount of operation given to the controlled object in a stepwise manner, and determines the dead time L and time constant slope R of the step response characteristic from the accompanying change in the amount of control. Then, in the process of increasing the controlled variable, when the controlled variable almost reaches the target value, the manipulated variable is decreased in steps, and the time from the point of decrease until the change in the controlled variable changes from increasing to decreasing in response. A correction dead time measurement means that obtains the correction dead time Ln as the correction dead time Ln, corrects the value of the obtained step response dead time L using the correction dead time Ln by simple averaging or weighted average processing, and calculates the value after this correction. Using the dead time Ls and the time constant slope R, find the proportional sensitivity Kp, integral time Ti, and differential time Td.
PID constant calculation means and the obtained PID constants Kp, Ti
and a main calculation means for guiding the controlled variable to the target value by PI or PID control using Td and Td.

FIG. 1 is a block diagram of a PID controller to which the present invention is applied. This controller 1 is a digital controller equipped with an AD converter, and its configuration is as follows: a main processing section 2;
It is roughly divided into a PID constant calculation section 3, a step response measurement section 4, and a correction dead time measurement section 5.

The main calculation unit 2 receives the target value r set as appropriate and the control amount C detected from the controlled object 6, calculates the deviation e(t) between the two, and calculates the PID
The above is calculated using the PID constants K _P , T _i , T _d obtained by the constant calculation unit 3 as described later and the above deviation e(t).
Calculate equation (2) and output the calculation result y as the manipulated variable. A changeover switch 7 is provided on the output side of the main calculation section 2, and in the illustrated normal usage state, the manipulated variable y output from the main calculation section 2 is applied to the controlled object 6. This performs well-known PID control.

Optimum adjustment of the PID constants K _P , T _i , and T _d is normally performed when the manipulated variable y for the controlled object 6 is zero and the controlled variable C is stable at a certain constant value C ₀ . Taking the temperature control of an electric furnace as an example, the heating current to the furnace is zero and the temperature of the furnace is stable at C ₀ . When a starting switch (not shown) is operated in such an initial stable state and a starting signal for optimum adjustment is given to the regulator 1, control and calculation for optimum adjustment are performed as follows.

When the start signal is given, the step response measuring section 4 switches the changeover switch 7 to the (b) side, and changes the predetermined amount of operation y _A output from the measurement section 4 to the changeover switch 8 in the state shown in the figure. At the same time, the control amount C from the controlled object 6 is introduced into the measuring section 4 through the changeover switch 7 . In other words, the manipulated variable is 0 for the controlled object 6.
A predetermined manipulated variable y _A is applied in a stepwise manner from the state of , and a controlled variable C that changes depending on the step response of the controlled object 6 to this step input is taken into the measuring section 4 .

Section I in FIG. 2a is a time chart of a step input to the controlled object 6, and section I in FIG. 2b is a general example of a step response of a multicapacity localization type process with dead time. When the controlled object 6 is a multicapacity localization type process with dead time, the control amount C introduced into the measuring section 4 changes from the initial value C ₀ so as to draw an S-shaped curve as shown in the figure. That is, the control amount C increases slowly at first, shows the maximum rate of increase at the middle point, and then increases slowly again.
The measuring section 4 measures this step response and detects the dead time L and the maximum slope _yAR .

As is well known, the maximum slope y _A R is the rate of change at the inflection point where the rate of change of the step response characteristic is maximum, and this is a value proportional to the step operation amount y _A , but its proportional constant R is called the time constant slope. Furthermore, as shown in Fig. 2a, the dead time L is determined by the tangent line drawn to the above inflection point of the step response curve from the time when the step input is applied to the initial value _C0.
is the time to the point where it intersects a straight line parallel to the time axis passing through .

The step response measurement unit 4 measures the step operation amount y _A
After the application of , the control amount C is sampled at a very small constant period, and its rate of change is sequentially checked to detect the maximum value _yAR of the rate of change. At the same time, the time T ₁ until the maximum value y _A R is generated and the control amount C ₁ at that time are detected. Then, the time constant slope R is calculated from the detected y _A R, and from the detected y _A R, T ₁ and C ₀ , the above dead time which has the relationship of C ₁ /T ₁ -L=y _A R is calculated. Calculate L. Furthermore, the time constant slope R and dead time L obtained as a result are supplied to the PID constant calculation section 3.

On the other hand, after R and L are detected by the measuring section 4 as described above, the correction dead time measuring section 5
Introducing the control amount C that increases further, the target value r
When the control amount C increases to a value r-α that is smaller than the target value r by a predetermined value (second
( _b )
The operation amount y _B (≪y _A ) output from the measuring section 5 is applied to the controlled object 6 via the switches 8 and 7. In other words, the amount of operation applied to the controlled object 6 is decreased stepwise from _yA to _yB . Note that this manipulated variable y _B is a value such that C ₂ ≪r−α, where C ₂ is the controlled variable when y _B is applied to the controlled object 6 for a long time and the system is stable. y _B =
0 is also fine.

Therefore, in response to the decrease in the manipulated variable from y _A to y _B , as shown in FIG. 2b, the controlled object 6
After a certain delay time due to the characteristics of , the change in the control amount C reverses from increasing to decreasing. Such a change in the control amount C appears near the target value r.

The measuring unit 5 calculates the time L o from the time t ₁ when the manipulated variable is changed to y _B to the time t ₂ when the change in the controlled variable C is reversed, that is, the time t ₂ when the peak value of the controlled variable _C appears. Detected as dead time L _o for correction. That is, the measurement unit 5 starts a time counting operation from time t ₁ and also starts a peak detection operation of the controlled amount C, and reaches the time t ₂ when the peak value of the controlled amount C is detected.
to stop the time counting operation. What is obtained as a result is the dead time L _o , and the measuring section 5 supplies the detected dead time L _o to the PID constant calculation section 3 .

The dead time L measured and detected in this way
As described later, _o is an element for calculating the corrected dead time L _S , and this dead time L _o
indicates the response delay of the controlled object 6 near the target value r, and this varies depending on the magnitude or height of the target value r. In other words, this dead time L _o is a value that reflects the characteristics of the controlled object 6 at each target value. This is the first aim of finding the dead time L _o . Also, this dead time L _o is
The previously determined dead time L represents the response delay when the manipulated variable is increased, whereas it represents the response delay when the manipulated variable is decreased. In other words, in general, when controlling the controlled object 6 on and off, the controlled object 6 is turned on to increase the amount of operation.
Normally, the characteristics of the system are different between the control state and the OFF control state, which is the opposite. Therefore, if you calculate the second dead time L _o as described above, this is the control state. Data reflecting the characteristics of the object 6 during cooling will be obtained. This is the second aim of finding the dead time L _o .

As mentioned above, following the dead time L, the dead time L
When _o is determined, the controller 1 calculates the
Furthermore, by repeating heating and cooling, preparations are made for fine-grained control depending on the control method adopted.

Next, the PID constant calculation section 3 calculates the average value L _S =L+Ln/2 of the dead time L given from the measurement section 4 and the dead time L _o given from the measurement section 5, and this is corrected. Let the dead time L _S be. Furthermore, the calculation unit 3 calculates PID constants K _P , T _i , T _d expressed by the following equations from the dead time L _S and the above-mentioned time constant slope R.

K _P =a/R・L _S , T _i =b・L _S , T _d =c・L _S ……(
3) This formula (3) is known as a theoretical formula for calculating the optimal value of the PID constant from the dead time of the step response and the slope of the time constant. is the value given.

PID constant K calculated as described above in calculation unit 3
_P , T _i and T _d are supplied to the main calculation section 2 and set as parameters for calculating the manipulated variable y from the deviation e(t) as described above, and the changeover switches 7 and 8 are returned to the normal state shown. be done. This completes the optimal adjustment control.

As mentioned above, in the case of the PID controller 1,
Regarding the actual controlled object 6, not only the dead time L and time constant slope R of the step response but also the response delay near the target value are detected as the dead time L _o for correction, and the average value L _S of L and L _o is detected. Since the PID constant is calculated based on the time constant slope R, even if the controlled object 6 includes nonlinear elements, the optimal PID constant for control near the actual target value can be determined, and this allows highly accurate PID Control can be realized. Moreover,
Optimal adjustment of the PID constant is automatically performed simply by operating a predetermined start switch, etc., so even if the target value is frequently changed, optimal control can always be achieved.

In the above embodiment, the average value of L and _Lo is the corrected dead time L _S , but an appropriate weighted average of L and _Lo may be used as L _S .

In addition, the main calculation unit 2 performs the calculation of equation (1) above.
In the case of PI control, the data on the differential time T _d is not required, and the process is almost the same as the above-mentioned PID control, so the explanation will be omitted.

Furthermore, in the above embodiment, the main calculation section 2, the PID constant calculation section 3, the step response measurement section 4,
The correction dead time measuring sections 5 are shown as independent circuit blocks, but this is for the convenience of explanation, and since this controller involves complex calculation operations, it is not implemented as an LSI. processor,
From the viewpoint of device cost, it is desirable to configure the regulator 1 mainly using a so-called microprocessor. That is, if a microprocessor is used, it is part of the main memory for the microprocessor.
By manipulating the necessary parts of RAM (Random Access Memory) with a program stored in ROM (Read Only Memory), which is also a part of main memory, each of the above parts 2, 3, 4, and 5 can perform the functions described above. Substantially equivalent functionality is obtained, which is obvious to those skilled in the art. However, in order to make it easier to understand the configuration when using a microcomputer, FIG. 3 shows a flowchart centered on the program for measuring the dead time L _o . In other words, this flowchart calculates the time from when the manipulated variable is changed from y _A to y _B until the controlled variable C reaches its peak by counting how many times the sampling timing pulse arrives during this time. This shows that L _S is obtained by setting this measurement result to the above-mentioned L _o , the PID constant is determined based on this L _S , and then the original PID control is performed. Note that a part of the RAM is allocated to the counter described above, and this counter and the part that stores the program that operates this counter work together to calculate the dead time for correction mentioned above. This forms a measuring section, which is consistent with the previous explanation.

Furthermore, in the previous embodiment, when the controlled variable C becomes a value lower than the target value r by a constant value α, the manipulated variable is changed from y _A to y _B
However, if the control amount C is expected to greatly exceed the target value,
In order to eliminate this problem, the timing of the control amount C relative to the target value r is predicted by measuring the change in the control amount C even after the maximum slope y _A R is detected.
What is necessary is just to switch from heating to cooling in accordance with this.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of one embodiment of the present invention, and FIG.
Figures a and 2b are timing charts for explaining the operation of optimal adjustment of PID constants, and Figure 3 is a flow chart showing a basic operation program when the present invention is realized by microprocessor control. DESCRIPTION OF SYMBOLS 1...Adjuster, 2...Main calculation unit, 6...Controlled object, C...Controlled amount, r...Target value, y...Manipulated amount.

Claims (1)

[Scope of Claims] 1. A step response measuring means for increasing a manipulated variable given to a controlled object in a stepwise manner and determining a dead time L and a time constant slope R of a step response characteristic from the accompanying change in the controlled variable. , during the process of increasing the controlled variable, when the controlled variable almost reaches the target value, the manipulated variable is decreased in steps,
A correction dead time measuring means for determining the time from the point of decrease until the change in the controlled variable shifts from increase to decrease in response to this as the correction dead time Ln, and a value of the determined step response dead time L. Corrected by simple averaging or weighted average processing using correction dead time Ln, and the dead time after this correction
Proportional sensitivity using Ls and the time constant slope R
PID to find Kp, integral time Ti and derivative time Td
A PI or PID controller comprising: a constant calculation means; and a main calculation means for guiding a controlled variable to a target value through PI or PID control using the determined PID constants Kp, Ti, and Td.