JPH05297908A - Adaptive controller - Google Patents

Abstract

PURPOSE:To effectively suppress variation in controlled variable in the case of the application of unknown disturbance to a process by improving the control performance and eliminating stationary deviation. CONSTITUTION:A system which is provided with a transfer function part 5 in parallel to the process 1 and makes an extended process controlled variable ya, obtained by adding its output and the controlled variable yp of the process 1, follow up the output ym of a standard model 5 feeds the deviation ez between the model output ym and controlled variable ya back through a variable gain. A 1st phase advance compensator 53 is provided on the upstream side of the variable gain and a 2nd phase advance compensator 74 is provided right before the process 1; and some parameters are varied in proportion to the integral value of the square root of the absolute value of the control deviation in a specified time zone. A 1st feedforward compensating means which removes the stationary deviation by using the input of the standard model and a 2nd feedforward compensating means which compensates the delay of the process by using a signal of the known disturbance are provided to learn the parameters on a on-line basis.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an adaptive control device applied to processes, machine products and the like.

[0002]

2. Description of the Related Art FIG. 3 shows a typical example of a conventional adaptive control apparatus applied to a process, a mechanical product or the like.

A process 1 inputs a known disturbance (d (t)) 2 and a manipulated variable (u (t)) 3 and outputs a controlled variable (yp (t)) 4. The control amount (yp (t)) 4 needs to follow the output ym (t) 6 of the reference model 5. In addition, normative model 5
The input um (t) 7 of is separately given as a command value. The second output xm (t) 8 of the reference model 5 is an intermediate variable of the reference model 5. Originally, control amount (yp (t)) 4
And the output ym (t) 6 are compared, and the manipulated variable (u (t)) 3 is moved so as to reduce the control deviation, but here, the transmission of Gb (s) is performed for stabilizing the control. The function part 9 is arranged in parallel with the process 1, and the control amount (yp (t)) 4 and Gb (s)
The output 10 of the transfer function unit 9 is added by the adder 11,
A deviation ez (t) 13 between the output ym (t) 6 and the ya (t) 12 when the so-called extended process control value is ya (t) 12.
Is calculated by the subtractor 14, and the manipulated variable (u (t)) 3 for reducing the deviation ez (t) 13 is calculated by the following method. The components for calculating the manipulated variable (u (t)) 3 are
It can be roughly divided into three. One of them is as follows.

The deviation ez (t) 13 is used as one side input of the multipliers 15 and 16, the output KIe (t) 18 of the integrator 17 described later is input to the other side of the multiplier 15, and the other side of the multiplier 16 is input. The output Kpe (t) 20 of the coefficient unit 19 described later is input. Then, the outputs of the multipliers 15 and 16 are supplied to the adder 21 and added.

Here, the output KIe (t) 18 of the integrator 17
Is input to the minus terminal of the subtractor 23 via the coefficient unit 22. The output of the subtractor 23 becomes the input of the integrator 17. The output of the coefficient unit 24 is input to the + terminal of the subtractor 23. The output of the multiplier 25 is input to both the coefficient units 19 and 24, and the deviation ez (t) 13 output from the subtractor 14 is input to both input terminals of the multiplier 25. Similarly, constituent elements for calculating the manipulated variable (u (t)) 3 will be described.

The second output xm (t) 8 of the reference model 5 is used as one side input of the multipliers 26 and 27, and the output KIx (t) 29 of the integrator 28 described later is input to the other side of the multiplier 26, An output Kpx (t) 31 of a coefficient unit 30 described later is input to the other side of the multiplier 27. Then, the outputs of the multipliers 26 and 27 are supplied to the adder 32 and added.

Here, the output KIx (t) 29 of the integrator 28
Is input to the minus terminal of the subtractor 34 via the coefficient unit 33. The output of the subtractor 34 becomes the input of the integrator 28. The output of the coefficient multiplier 35 is input to the + terminal of the subtractor 34. The output of the multiplier 36 is input to both the coefficient units 30 and 35, and the multiplier 36 outputs the deviation ez (t) 13 output from the subtractor 14 and the second output xm (t) of the reference model 5.
8 is input.

The third component is that the input um (t) 7 of the reference model 5 is used as one side input of the multipliers 37 and 38, and the multiplier 3
The output KIu (t) 40 of the integrator 39, which will be described later, is input to the other side of 7
Enter pu (t) 42. Then, the outputs of the multipliers 37 and 38 are supplied to the adder 43 and added.

Here, the output KIu (t) 40 of the integrator 39
Is input to the minus terminal of the subtractor 45 via the coefficient unit 44. The output of the subtractor 45 becomes the input of the integrator 39. The output of the coefficient multiplier 46 is input to the + terminal of the subtractor 45. The output of the multiplier 47 is input to both the coefficient units 41 and 46, and the deviation ez (t) 13 output from the subtractor 14 and the input um (t) 7 of the reference model 5 are input to the multiplier 47.
Is the input. The adders 21 and 32 of each of the above components
Each output of 43 is input to the adder 48, and this adder 4
The output of 8 is the manipulated variable (u (t)) 3 for process 1.

The transfer function unit 9 of Gb (s) may be a first-order lag element in a simple case, and the reference model 5 similarly outputs the output of the first-order lag element in a simple case ym (t) 6. Then, the value corresponding to the differential value of the output ym (t) may be used as the second output xm (t) 8. Here, the integrators 17, 28,
Each output of 39 must be limited to a positive value of zero or more as a minimum value. Also, the coefficient unit 19,
It is necessary to limit each output of 30 and 41 to a positive value of zero or more as a minimum value.

[0011]

The above-mentioned conventional adaptive control device is basically a proportional operation, and there is a limit to improvement of control performance. Moreover, the steady-state deviation remains, which is not preferable.

Further, the above-mentioned conventional apparatus is effective in the case where the control amount of the process is made to follow the output of the reference model, but when a known disturbance is added to the process, the feedforward compensator suppresses the fluctuation of the control amount. Since it is not performed, there is a problem that high control performance cannot be obtained.

The present invention has been made in view of the above circumstances, and is capable of improving control performance, eliminating steady-state deviations, and effectively suppressing fluctuations in the control amount when a known disturbance is added to the process. The purpose is to provide.

[0014]

An adaptive control apparatus according to the present invention is obtained by providing a transfer function unit having a manipulated variable as an input in parallel with a process, and adding the output and a control amount yp of the process. In the system aiming to make the expanded control value ya of the process follow the output ym of the reference model,
A means for feeding back the deviation ez obtained by subtracting the expanded control value ya of the process from the output ym of the reference model, the first phase lead compensator arranged upstream of the feedback loop, and the control value yp of the process. Of 2
A second phase lead compensator for feeding back the differential value to suppress the oscillatory behavior of the process;
Means for changing some parameters of the phase advance compensator in proportion to the integral value of the square root of the absolute value of the control deviation in the specified time zone, and the addition of the input of the reference model in the main loop of the feedback control. A first feedforward compensating means for inputting to the point to remove the steady deviation;
Control amount yp of the above process and output ym of the above reference model
And a second feedforward compensating means for compensating for the delay of the process by using the signal of the known disturbance so as to minimize the difference between the first and second feedforward compensating means. And online learning of the parameters of the second feedforward compensating means so that the difference between the expanded control value ya of the process and the output ym of the reference model is advanced by the phase advance compensator. And a second learning means for performing the learning.

[0015]

By providing the phase lead compensators at two locations as described above, the manipulated variable that compensates for the process delay can be obtained. Further, by arranging the addition point when the input of the reference model is used as the feedforward compensation, not in the bypass loop but in the main loop, the steady deviation can be minimized in addition to the function of the feedforward compensation.

Further, the value e ^{* obtained} by advancing the deviation ez in the feedforward compensation of the known disturbance by the first phase advance compensator ^. A modification to z is made so that the process delay results in effective feedforward compensation.

Further, the operation amount capable of suppressing the vibration when the process exhibits a vibrational behavior can be obtained by the second phase lead compensator. This is the controlled variable yp of the process
This is because the value obtained by differentiating the second order is fed back.
Then, the value corresponding to the differential time is automatically corrected to an appropriate value according to the magnitude of the control deviation ey.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. The circuit configuration of the device of the present invention is shown in block diagrams in FIGS. The same parts as those of the conventional device shown in FIG. 3 are designated by the same reference numerals, and detailed description thereof will be omitted.

In FIG. 1, a process 1 inputs a known disturbance (d (t)) 2 and a manipulated variable (u (t)) 3 and outputs a controlled variable (yp (t)) 4. Where the controlled variable (yp (t))
4 must follow the output ym (t) 6 of the reference model 5. The input um (t) 7 of the reference model 5 is separately given as a command value. Originally, the control amount (yp (t)) 4
And the output ym (t) 6 are compared by the subtractor 49, and the manipulated variable (u (t)) 3 so that the control deviation (ey (t)) 50 is reduced.
, But here, for control stabilization, k
The output ym (t) 6 and ya (t) 6 and ya (t) when the value of the so-called extended process control amount obtained by adding the output 52 of the transfer function unit 51 of 1 / (1 + T1 S) by the adder 11 is ya (t) 12 t)
The deviation ez (t) 13 of 12 is calculated by the subtractor 14, and the manipulated variable (u (t)) 3 for reducing the deviation ez (t) 13 is also calculated in parallel by the following method. The components for calculating the manipulated variable (u (t)) 3 are roughly divided into three. One of them is as follows.

The deviation ez (t) 13 output from the subtractor 14
A value e ^* obtained via the first phase lead compensator 53 z
(t) 54 is used as one side input of the multipliers 15 and 16, and the multiplier 1
On the other side of 5, the output KIe of the integrator 55 with upper and lower limits described later
(t) 18 is input, and the output Kpe (t) 20 of the coefficient unit with upper and lower limits 56 described later is input to the other side of the multiplier 16. Here, the first phase lead compensator 53 supplies the deviation ez (t) 13 to the adder 57 and the differentiator with filter 58, and the output of this differentiator with filter 58 is added via the coefficient unit 59. 5
Type in 7. The output of the adder 57 is taken out as the output of the first phase advance compensator 53. The outputs of the multipliers 15 and 16 are supplied to the adder 60.

As shown in FIG. 2, the output KIe (t) 18 of the integrator 55 with upper and lower limits is supplied to the − of the subtractor 23 via the coefficient unit 22.
Input to the terminal. The output of the subtractor 23 becomes the input of the integrator 55 with upper and lower limits. Here, the integrator with upper and lower limits is
It means that the output of the integrator is limited by the specified upper and lower limits. The same applies to the coefficient unit with upper and lower limits. Then, the output of the coefficient unit 24 is input to the + terminal of the subtractor 23. The output of the multiplier 25 is input to both the coefficient unit with upper and lower limits 56 and the coefficient unit 24, and the multiplier 25 receives the deviation ez (t) 13 and the value e ^*. z (t) 54 becomes an input. Similarly, the manipulated variable (u (t)
) The components for calculating 3 will be described.

In FIG. 1, the input um (t) of the reference model 5
7 is one-sided input of the multiplier 37, the output KIu (t) 40 of the integrator 39 described later is input to the other side of the multiplier 37, and the output Kpu (of the coefficient unit 41 described later is to the other side of the multiplier 38. t) 4
Enter 2. The outputs of the multipliers 37 and 38 are
It is supplied to the adder 61. In addition, the integrator 39 shown in FIG.
The output of the coefficient unit 46 is input to. Coefficient unit 41, 4
The output of the multiplier 47 is input to both 6 and the multiplier 47
The control deviation ey (t) 50 and the input um (t) 7 of the reference model 5 are input.

As the third component, as shown in FIG. 1, a known disturbance (d (t)) 2 is used as one side input of the multipliers 62 and 63, and the other side of the multiplier 62 is connected to an integrator 64 to be described later. Output Kid
(t) 65 is input, and an output Kpd (t) 67 of a coefficient unit 66 described later is input to the other side of the multiplier 63. And the multiplier 6
The respective outputs of 2, 63 are supplied to the adder 60.

Here, the output K id of the integrator 64 shown in FIG.
The (t) 65 is input to the-terminal of the subtractor 69 via the coefficient unit 68. The output of the subtractor 69 is input to the integrator 64. The output of the coefficient unit 70 is input to the + terminal of the subtractor 69. The output of the multiplier 71 is input to both the inputs of the coefficient units 66 and 70, and the multiplier 71 outputs the value e ^*. z (t) 54 and known disturbance (d (t)) 2 are input.

The output up (t) 72 of the adder 60 shown in FIG. 1 obtained by each of the above components is input to the adder 61 and the transfer function unit 51. The output 73 of the adder 61 and the output 82 of the second phase lead compensator 74, which will be described later, are added to the adder 7
5 is added and the output becomes the manipulated variable (u (t)) 3.

In the second phase lead compensator 74, the control amount (yp (t)) 4 output from the process 1 is input to one side of the multiplier 76 via the differentiator with filter 75, and the other is described later. The output Kb (t) 78 of the quantization element 77 of is input. The output of the multiplier 76 is connected and input to the-terminal of the subtractor 79, and the output 73 of the adder 61 is input to the + terminal. Then, the output of the subtractor 79 is input to the multiplier 81 via the differentiator with filter 80, and is multiplied by the output Kb (t) 78 of the quantization element 77, which will be described later, and the output thereof advances by the second phase advance. It becomes the output 82 of the compensator 74.

Next, a method of calculating the output Kb (t) 78 will be described. Control deviation ey (t) 50 output from subtractor 49
Is input to the absolute value generator 83 shown in FIG. 2, and its output is input to the square root generator 84. The output of the square root generator 84 is applied to the + terminal of the subtractor 85 and compared with the output of the coefficient multiplier 86. The output of the subtractor 85 is integrated by the integrator 87 and then supplied to the coefficient unit 86, the subtractor 88, and the dead time generator 89. The output of the dead time generator 89 is input to the minus terminal of the subtractor 88. The output of the subtractor 88 becomes the input of the quantization element 77 via the coefficient unit 90 with upper and lower limits, and the output of this quantization element 77 becomes Kb (t) 78. By providing the first phase lead compensator 53 and the second phase lead compensator 74 as described above, the manipulated variable u (t) that compensates for the process delay can be obtained.

Further, since the addition point when the input um (t) 7 of the reference model 5 is used as the feedforward compensation is arranged not in the bypass loop but in the main loop, in addition to the function of the feedforward compensation, The circuit configuration can minimize the steady-state deviation. That is, as long as there is a control deviation, the operation amount correction operation is performed by the integration function, and the steady deviation can be eliminated.

Further, the feedforward compensation of the known disturbance (d (t)) 2 is performed by using the deviation ez (t) 13 as the first phase advance compensator 5.
The correction is made to the value e * z (t) advanced in 3. As a result, the process delay becomes effective feedforward compensation, and the control performance can be improved.

Further, the operation amount capable of suppressing the vibration when the process shows a vibrational behavior is the second phase advance compensator 74.
Determined by. This is because the value of the control amount (yp (t)) 4 which is the second-order differential (all the differentials here mean that with a filter) is fed back. Then, the value corresponding to the differential time is automatically corrected to an appropriate value according to the magnitude of the control deviation ey (t). The method of obtaining the value corresponding to the appropriate differential time is obtained by quantizing the integral value of the square root of the absolute value of the control deviation ey (t) in the designated time zone L. The reason for taking the square root is to reduce the influence of the peak value of the control deviation. This can dampen the oscillatory behavior of the process. Here, L means dead time, and the value of L must be specified. Further, the quantizing element is provided in order to avoid the small movement of Kb (t) 78.

[0031]

As described in detail above, according to the present invention, the first phase lead compensator and the second phase lead compensator are used.
High control performance can be obtained. In addition, since appropriate parameters for the second phase lead compensator are obtained online, high control performance can always be obtained even if the process changes over time.

Further, by arranging the addition point in the main loop when the input um (t) of the reference model is used as feedforward compensation, and by making the control deviation itself the minimum value, the effect of feedforward compensation is obtained. It is possible to simultaneously improve the two points, that is, the removal of the steady deviation.

Further, by providing the feedforward compensator based on the known disturbance (d (t)), the known disturbance (d
(t)) can improve the control performance when it fluctuates, and the output e ^* of the first phase lead compensator is used as an index of the parameter at that time ^. By taking the method of minimizing z (t), the control performance can be improved significantly. Further, by providing a second phase lead compensator and feeding back the second derivative (with a filter), the behavior when the process becomes oscillatory can be improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing a main part of an adaptive control device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing another part of the adaptive control device according to the embodiment.

FIG. 3 is a block diagram showing a conventional adaptive control device.

[Explanation of Codes] 1 ... Process, 2 ... Known disturbance (d (t)), 3 ... Manipulation amount (u (t)), 4 ... Control amount (yp (t)), 5 ... Reference model,
9 ... Transfer function part 11, 21, 32, 43, 48 ... Adder, 12 ... Extended process control amount ya (t), 13 ... Deviation ez (t), 14, 23, 34, 45, 49 ... Subtractor, 1
5,16,25,25,27,36,37,38,47
... multiplier, 17, 28, 39 ... integrator, 18 ... output KIe
(t), 19, 22, 24, 30, 33, 35, 41, 4
4, 46, 59 ... Coefficient unit, 20 ... Output Kpe (t), 29 ...
Output KIx (t), 31 ... Output Kpx (t), 40 ... Output KIu
(t), 42 ... Output Kpu (t), 50 ... Control deviation ey (t), 5
1 ... Transfer function part, 53 ... First phase lead compensator, 54 ...
e ^* z (t), 55 ... Integrator with upper / lower limit, 56 ... Coefficient unit with upper / lower limit, 57, 60, 61 ... Adder, 58 ... Differentiator with filter, 62, 63, 71, 81 ... Multiplier, 64, 87 ... integrator, 65 ... output kid (t), 66, 68, 70, 86 ...
Coefficient unit, 67 ... Output Kpd (t), 69, 76, 79, 8
5, 88 ... Subtractor, 72 ... Output up (t), 74 ... Second phase advance compensator, 75 ... Filter differentiator, 77 ... Quantization element, 78 ... Output Kb (t), 80 ... With filter Differentiator, 8
3 ... Absolute value generator, 84 ... Square root generator, 89 ... Dead time generator, 90 ... Coefficient unit with upper and lower limits.

Claims (1)

[Claims] 1. A transfer function unit for inputting a manipulated variable is provided in parallel with a process, and its output and process controlled variable yp.
In a system aiming to make the extended process control amount ya obtained by adding the reference model output ym, the deviation ez obtained by subtracting the extended process control amount ya from the reference model output ym is given. A means for feeding back, a first phase lead compensator arranged on the upstream side of the feedback loop, and a second-order differential value of the controlled variable yp of the process are fed back to suppress the oscillatory behavior of the process. And a means for changing some of the parameters of the first and second phase advance compensators in proportion to the integral value of the square root of the absolute value of the control deviation in the designated time zone, and the above norm. 1st input of the model is input to the addition point provided in the main loop of the feedback control to remove the steady deviation
Feedforward compensation means of the above, the control amount yp of the above process and the output ym of the above reference model
A first learning means for online learning the parameters of the first feedforward compensating means so as to minimize the difference between the first and second feedforward compensating means, and a second learning means for compensating the process delay using the signal of the known disturbance.
And the second feedforward compensator for minimizing the difference between the expanded process control amount ya and the reference model output ym through the phase advance compensator. And a second learning means for online learning of the parameter (1).