JPH06131006A - Servo-controller - Google Patents

Abstract

PURPOSE:To improve a servo-follow-up property at the time of controlling a control object such as a servomotor, etc., by deriving automatically a parameter for compensating a control system. CONSTITUTION:Each deviation to a command value taken U [k] to a servomotor 1 and a controlled variable train Y [k] of the servomotor 1 corresponding to this command value train is derived repeatedly, and a compensation input train Uc [k] for converging this deviation to zero is derived by a compensation value learning function. Subsequently, based on this compensation input train Uc [k], each parameter for compensating feedback control is derived by a compensation determining function. Accordingly, at the time of deriving a manipulated variable of the servomotor 1, the manipulated variable is compensated, based on each parameter derived by the compensation determining function and given to the servomotor 1, and by this compensation, the servo-follow-up property is improved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a servo control device having improved servo followability when controlling a controlled object such as a servo motor.

[0002]

2. Description of the Related Art For servo control of a servo motor or the like, there is one that employs learning control, PID auto tuning, a neural network or the like to improve servo followability. Among them, PID auto tuning is
The feedback control system is used to adjust the gain, that is, each gain of proportional, integral and derivative (PID).

FIG. 4 is a block diagram of a feedback control system adopting such PID auto tuning. This control system controls the servo motor 1,
The command value is given to the adder 2 and a control amount fed back from the servo motor 1 is given to the adder 2, and the adder 2 outputs a deviation between the command value and the control amount. The output of the adder 2 is sent to the adder 3 through the proportional gain Kp.

At the same time, the output of the adder 2 is sent to the adder 3 through the one-sample delay element Z ^-1 and the integral gain Ki, and the control amount of the servo motor 1 is one sample of the delay element Z ^-1 and. It is sent to the adder 3 through the differential gain Kd. Then, the output of the adder 3 is given to the servo motor 1 as an operation amount. In such a control system, proportional, integral and derivative gains Kp, Ki and K are provided in order to improve the followability of the servo motor 1 to the command value.
d is being adjusted.

However, the feedback control system alone has a limit in improving the followability. For example, as shown in FIG. 5, the control amount B does not follow the command value A.

In order to improve the followability, a feedforward compensator 4 is connected as shown in FIG. The compensator 4 sends the command value to the adder 5 through its proportional gain FKp, and also sends the command value to the adder 5 through the one-sample delay element Z ^-1 and the differential gain FKd, and adds the output of this adder 5 It is supposed to be given to vessel 3.

However, these gains FKp, F
Kd and the delay element Z ^-1 must be obtained by identifying the servo motor 1, and these gains FK
Even if p, FKd, etc. are obtained, the load on the servo motor 1 will fluctuate.
p and FKd must be adjusted. Such adjustment of the gains FKp and FKd is extremely difficult because the algorithm is complicated.

On the other hand, in the above learning control, in order to obtain the optimum control amount sequence for the command value sequence given in time series, the compensation amount according to the fluctuation of the servo motor 1 is generated by learning. In this case, the compensation amount is generated by learning for each different command value. For this reason, a large capacity memory is required to store the compensation amount,
Besides, it is weak against disturbance.

In the case of using a neural network, the structure of the neural network is unknown.
It is unknown what kind of output is obtained when the input value is changed. Therefore, it cannot be used unless the layer structure of the neural network is determined in advance.

[0010]

As described above, learning control requires a large-capacity memory, and the neural network must determine its layer structure in advance. Further, in PID auto tuning, followability with respect to the command value is poor.

SUMMARY OF THE INVENTION It is therefore an object of the present invention to provide a servo control device capable of improving servo followability when automatically controlling parameters for compensating a control system and controlling a controlled object such as a servo motor. .

[0012]

DISCLOSURE OF THE INVENTION The present invention provides a servo controller for processing a command value by each gain of a control system to obtain a manipulated variable, and giving the manipulated variable to a controlled object for control.

Compensation value for repeatedly obtaining each deviation between the command value sequence input in time series and the control amount sequence output from the controlled object corresponding to this command value sequence and for obtaining the compensation input sequence that converges this deviation to zero. The above-mentioned object is provided with a learning means and a compensation determination means for determining each parameter for compensating the control system based on the compensation input sequence obtained by the compensation value learning means and compensating the control system based on these parameters. This is a servo control device that aims to achieve

[0014]

By providing such means, each deviation between the command value sequence for the control object input in time series and the control amount sequence output from the control object corresponding to this command value sequence is repeatedly obtained, A compensation input sequence that converges this deviation to zero is obtained by the compensation value learning means. Then, each parameter for compensating the control system is obtained by the compensation determining means based on this compensation input sequence.

Therefore, when the command value is processed by each gain of the control system to obtain the operation amount, the operation amount is compensated based on each parameter obtained by the compensation determining means and is given to the controlled object. This compensation improves the servo followability.

[0016]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. The same parts as those in FIG. 4 are designated by the same reference numerals, and detailed description thereof will be omitted.

FIG. 1 is an overall block diagram of the servo controller. A feedforward compensator (Filt) 10 is connected between the input end of the adder 2 and the adder 3. The compensator 10 has a compensation value learning function and a compensation determination function.

Of these, the compensation value learning function is a command value sequence U [k] input in time series and a control amount sequence Y [k] output from the servomotor 1 corresponding to this command value sequence U [k]. Each deviation is repeatedly obtained, and the compensation input sequence Uc [k] that converges this deviation to zero is obtained. Note that k = 1 to
n.

FIG. 2 shows a block diagram when the compensation value learning function is activated. That is, the feedforward compensator 10 has the delay element Z ^{−1 of} one sample and the learning gain α. Therefore, the compensation input sequence Uc [k] that is the output of the compensator 10 is expressed by the following equation. Uc ⁱ [k] = α {(Y [k] -U [k])-(Y [k-1] -U [k-1])} + Uc ^i-1 [K] (1) Then, the compensation input sequence Uc [k] is obtained as follows.

First, Uc is determined by the first trial (i = 1).
¹ Find [k]. In this case, the command value sequence U [k] is given, for example, for 20 seconds, and the compensation input sequence Uc ^{1 at} that time is given. [K]
(K = 1 to n) is calculated and stored.

Next, the same trial as above is repeated i times.
Given, compensation input sequence Uc at that timeⁱ [K] (k = 1 to
n) is found and stored. In other words, in the second trial
Is Uc² [K] Uc in the third trial³ [K], ...
Ucⁱ [K] is sequentially obtained for each trial, and the compensation input sequence Uc
Learn [k].

Further, the compensation determination function is for obtaining each parameter for compensating the feedback control system based on the compensation input sequence Uc [k] obtained by the compensation value learning function. That is, Uc [k] = Filt (Z ⁻¹ ) · U [k] (2) Filt (Z ⁻¹ ) is a time series model using Uc [k] and U [k], for example, AR. It can be obtained as a model or an ARMA model. For example, the formula (2) is represented by the following formula, Uc [k] = {( b1 · Z 2 + B2 · Z + b3) ÷ (Z ² + A1.Z + a2)} U [k] (3) and if there is a relation of the following equation, Uc [k] + a1 Uc [k-1] + a2 Uc [k-2] = b1 Uc [k] + b2 Uc [ k-1] + b3 Uc [k-2] (4) Each parameter a1, a2, b in these equations (3) and (4)
It becomes the calculation of 1, b2 and b3. Next, the operation of the apparatus configured as described above will be described.

(1) First, in order to determine each parameter of the feedforward compensator 10 for the servomotor 1, as shown in FIG. 2, the feed is performed between the adders 2 and 3 of the feedback system having the proportional gain Kp. The forward compensator 10 is connected. It should be noted that this connection may be performed by software processing.

In this state, the command value sequence U [k] is given as the first trial (i = 1), for example, for 20 seconds, and the compensation input sequence Uc ^{1 at} that time is given. [K] (k = 1 to n) is calculated and stored.

Then, the same trial as above is repeated i times.
Given, compensation input sequence Uc at that timeⁱ [K] (k = 1 to
n) is found and stored. In other words, in the second trial
Is Uc² [K] Uc in the third trial³ [K], ...
Ucⁱ [K] is sequentially obtained for each trial, and the compensation input sequence Uc
Learn [k].

Next, the compensation determination function is performed by the parameters a1, a2, b1, b2, given by the above equation (3), for example.
If it has b3, the parameters a1, a2, b1, b2 and b3 are determined by calculating the above equations (3) and (4) based on the compensation input sequence Uc [k].

(2) Next, the feedforward compensator 10
Is connected to the feedback control system of the servomotor 1 as shown in FIG. Note that this connection may be performed by software processing as in the above.

In such a feedback control system, the command value sequence U [k] is given to the adder 2 and the control amount sequence Y [k] fed back from the servomotor 1 is given to the adder 2, Command value sequence U from adder 2
The deviation between [k] and the control amount sequence Y [k] is output. The output of the adder 2 is sent to the adder 3 through the proportional gain Kp.

At the same time, the output of the adder 2 is sent to the adder 3 through a delay element Z ^{-1 of} 1 sample and an integral gain Ki, and the control amount sequence Y [k] of the servomotor 1 is delayed by 1 sample. It is sent to the adder 3 through the element Z ^-1 and the differential gain Kd.

On the other hand, when the command value sequence U [k] is sent to the feedforward compensator 10, the compensator 10
Is the compensation input value sequence Uc [k] corresponding to the command value sequence U [k]
Is given to the adder 3. Then, the output of the adder 3 is given to the servo motor 1 as an operation amount. As a result, the control amount sequence Y [k] of the servomotor 1 has a good followability with respect to the command value sequence U [k] as shown in FIG.

As described above, in the above embodiment, each deviation between the command value sequence U [k] and the control amount sequence Y [k] of the servomotor 1 corresponding to this command value sequence U [k] is repeatedly obtained. , A compensation input sequence Uc [k] that converges this deviation to zero, and each parameter a1, a2, ... Of the feedforward compensator 10 that compensates the control of the servomotor 1 based on this compensation input sequence Uc [k]. , B3 are obtained, by applying the compensator 10 to the feedback control system, the servo motor 1 for the command value sequence U [k] is obtained.
The controllability of the control amount sequence Y [k] can be improved.

Further, each parameter a1, a2, ..., B3 in the feedforward compensator 10 can be automatically obtained by trying several times. Therefore, even if the servo motor 1 to be controlled is changed to another servo motor having different characteristics, the parameters a1, a2, ..., B3 for the servo motor can be automatically obtained. This makes it possible to support various servo motors.

Further, if the pattern of the command value sequence U [k] is temporally scaled up / down, the command value sequence U [k].
The feedforward compensator 10 can be applied to each pattern without changing each parameter. It should be noted that the present invention is not limited to the above-described embodiment, and may be modified within the scope of the invention.

For example, the structure of the feedforward compensator 10 is determined in advance, and the repetitive command value is set to the servo motor 1
, And each parameter of the feedforward compensator 10 may be optimized so that the deviation between the command value and the control amount at that time becomes zero.

[0035]

As described above in detail, according to the present invention, the servo control capable of improving the servo followability when automatically controlling the parameters for compensating the control system and controlling the controlled object such as the servo motor is achieved. A device can be provided.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of a servo control device according to the present invention.

FIG. 2 is a block configuration diagram showing a compensation learning function of a feedforward compensator in the same apparatus.

FIG. 3 is a diagram showing the followability of the apparatus.

FIG. 4 is a block diagram of a conventional device.

FIG. 5 is a diagram showing the followability of the apparatus.

FIG. 6 is a block diagram of a conventional feedback control system.

[Explanation of symbols]

1 ... Servo motor, 2, 3 ... Adder, 10 ... Feedforward compensator, Z ^-1 ... Delay element, α ... Learning gain.

Claims (1)

[Claims] 1. A servo control device for processing a command value by each gain of a control system to obtain a manipulated variable, and giving the manipulated variable to a control object to control the command value sequence input in time series and Compensation value learning means for repeatedly obtaining each deviation from the control amount sequence output from the controlled object corresponding to the command value sequence and obtaining a compensation input sequence that converges this deviation to zero, and the compensation value learning means And a compensation determining unit for compensating the control system based on these parameters.