JPH05165504A - Learning control system by incremental value operation - Google Patents

Abstract

PURPOSE:To curtail a memory, and to shorten an arithmetic time by inputting an incremental value of a target command and an incremental value of an output of a control object, and outputting a correction command or its incremental value to the controlled system. CONSTITUTION:At the time (i) of a k-th trial time, an incremental value DELTAek (i) of a follow-up deviation and its integral value ek (i) are derived by a difference between an incremental value DELTAr (i) of a target command and an incremental value DELTAyk (i) of an output of a control object. Also, an incremental value {DELTAek (i+1), ..., DELTAek (i+M)} of a deviation to the future of M steps is predicted by the incremental value DELTAek (i) of the deviation, an incremental value of a deviation at the time of the previous trial, an incremental value of a correction amount of the past and the present time, and information related to a dynamic characteristic of the control object. Subsequently, an incremental value DELTAsigmak (i) of the correction amount of the present time is determined so that an evaluation function shown by an expression I becomes minimum, and an incremental value DELTAUk (i) of a correction command is derived by DELTAUk (i)=DELTAk-1 (i)+DELTAsigmak (i). In such a way, a necessary memory can be curtailed, and also, an arithmetic time can be shortened.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control system for machine tools, robots and the like that repeatedly operate.

[0002]

2. Description of the Related Art As a method for designing a learning control system with respect to a repetitive target value, there is a method proposed by the present applicant in Japanese Patent Laid-Open No. 1-237701. In the first invention of this system,
Repeat the operation for the same target value, predict the future deviation based on the deviation at the current and previous trials, the increment value of the current and past correction amount, and the step response of the controlled object,
Since the weighted sum of squares of the predicted value is used as the evaluation function and the control input is corrected so that the evaluation function becomes the minimum, the target value and the output eventually match and a highly accurate follow-up operation is realized. It Further, the applicant of the present invention, as a method of shortening the number of sampling points of the step response of the control system in the above-mentioned design method, samples only the first few step responses, and thereafter, the difference value of the step response is a constant damping ratio. A method that approximates the decrease is proposed in Japanese Patent Application No. 3-173091. Further, the present applicant, in Japanese Patent Application No. 2-196940, with respect to the position control system of the motor, insert a learning controller inside the position loop, and its input is a value that is a position tracking deviation or a constant multiple thereof, Predict the future deviation, and use the correction command u so that the weighted sum of squares of the predicted value is minimized.
A method is proposed in which (i) is corrected and the corrected speed command is input to the speed controller.

[0003]

However, in these methods, since the values of the target command, the output and deviation of the controlled object, the correction command, etc. are used as they are, the bit length of the data becomes long and the memory There was a problem such as increase of calculation time. Therefore, an object of the present invention is to reduce the required memory and the calculation time.

[0004]

SUMMARY OF THE INVENTION In order to solve the above problems, the present invention provides an incremental value of the target command and an incremental value of the output of the controlled object so that the output of the controlled object follows the target command that repeats the same pattern. In the control system that inputs the correction command or the increment value thereof to the control target, at the time i at the time of the k-th trial, the increment value Δr (i) of the target command and the increment value Δy _k of the output of the control target. The incremental value Δe _k (i) of the tracking deviation and its integrated value e _k (i) are obtained from the difference from (i), and the incremental value Δe _k (i) of the deviation and the previous trial value Based on the incremental value of the deviation, the incremental values of the correction amounts at the past and present times, and the information on the dynamic characteristics of the controlled object, M
Incremental value of deviation to step future {Δe _k (i + 1), Δe _k
(i + 2), ..., Δe _k (i + M)} is predicted and the evaluation function

[0005]

[Equation 2]

In order to minimize the correction amount, the increment value Δσ _k (i) of the correction amount at the current time is determined, and the correction command is given by Δu _k (i) = Δu _k-1 (i) + Δσ _k (i). It is characterized in that the increment value Δu _k (i) of is obtained.

[0007]

By the above means, a learning control system that operates with less memory and shorter calculation time is constructed.

[0008]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention can be applied to any of the systems described in the prior art or to other systems.
Here, a specific embodiment will be described with reference to FIG. 1 when applied to the first invention of JP-A-1-237701. In the figure, 1 is a command generator, which generates an increment value Δr (i) of the target command value r (i) at the current time i. 2 is a subtracter, which is a target command increment value Δr (i) and an output increment value y
and outputs a _k deviation increment Δe _k of the (i) (i). 3, a constant _{Q, A 1, A 2,} ..., A M, B, g 1, g 2, ..., stores g _N memory 4, the current time i from the time i at the previous trial
Memory for storing the deviation increment value up to the present time i, 5 is a memory for storing the correction amount increment value from N sampling past to the present time i, and 6 is a correction command from the time i at the time of the previous trial to the present time i. It is a memory for storing the increment value. Reference numeral 7 is an integrator for obtaining the current deviation e _k (i). 8 is an arithmetic unit,

[0009]

[Equation 3]

By this calculation, the correction amount increment value Δσ _k (i) at the time i is calculated. Further, 9 adds the current correction amount increment value Δσ _k (i) and the correction command increment value Δu _k-1 (i) at the time i at the time of the previous trial to obtain the current correction command increment value Δu _k ( It is an adder that outputs i). Reference numeral 11 denotes a motor and its position control system, which receives a correction command increment value Δu _k (i) as an increment value of a target position command and outputs an actual position y _k (i) of the motor. Reference numeral 10 is a differentiator for obtaining the increment value Δy _k (i) of the motor position. Here, the formula (1) is derived. At the time i at the time of the k-th trial, the deviation e _k (i + m) in the m step future is predicted by the following equation.

[0011]

[Equation 4]

Here, h _j (j = 1,2, ..., N) is the difference value (( _j) of the sampled value H _j of the output response of the motor position control system when a unit step is inserted as the target position command. h _j = H
_j − H _j−1 ), and N is chosen so that the response is well settled, ie h _n = 0 (n> N).
From equation (2), the deviation increment Δe _k (i + m) in the future m steps
Is

[0013]

[Equation 5]

Is predicted. However, it is assumed that the future correction amount increment value Δσ _k (i + 1) = Δσ _k (i + 2) = ... = 0. Now the evaluation function

[0015]

[Equation 6]

The increment value Δσ _k (i) of the correction amount at the current time is determined so that Therefore, ∂J / ∂Δ
From σ _k (i) = 0, Δσ _k that minimizes the evaluation function equation (4)
(i) is given by the above equation (1). However, each constant,
_{q m, Q, A m,} B, g n is given by the following equation.

[0017]

[Equation 7]

When the control target is not the command increment value but a command is input, the integral value u _k (i) of the output Δu _k (i) of the adder 9 is calculated and input to the control target. good. Here, the present invention is described in Japanese Patent Laid-Open No. 1-237701
The invention is applied to the invention, but in the other methods described in the prior art, or in other methods, the future deviation increment value is calculated by the deviation increment value and the correction amount as shown in Expression (3). The present invention can be applied by predicting the increment value and determining the current increment value of the correction amount so as to minimize the evaluation function expression (4).

[0019]

As described above, according to the present invention, it is possible to realize a learning control system which operates with a smaller memory and a shorter calculation time.

[Brief description of drawings]

FIG. 1 is a diagram showing a specific embodiment of the present invention.

[Explanation of symbols]

1 command generator 2 subtractor 3 constant Q, q ₁ , q ₂ , ..., q _M , g ₁ , g ₂ , ..., g _N memory 4 memory for storing deviation increment value 5 correction amount increment value Memory to be stored 6 Memory for storing correction command increment value 7 Integrator 8 Computing unit 9 Adder 10 Difference unit 11 Motor and its position control system

Claims (2)

[Claims] 1. A control for inputting an increment value of a target command and an increment value of an output of a control target, and outputting an increment value of a correction command to a control target so that the output of the control target follows the target command repeating the same pattern. Time i at the k-th trial in the system
In accordance with the difference between the incremental value Δr (i) of the target command and the incremental value Δy _k (i) of the output of the controlled object, the incremental value Δ of the tracking deviation is
e _k (i) and its integrated value e _k (i) are obtained, and further, the deviation increment value Δe _k (i), the deviation increment value at the previous trial, and the past and present time correction amounts. The increment value of
Based on the information on the dynamic characteristics of the controlled object, the increment value of the deviation to the M step future {Δe _k (i + 1), Δe _k (i + 2), ...,
Δe _k (i + M)} is predicted, and the evaluation function So that is minimized, the incremental value Δσ of the correction amount at the current time
_{The characteristic is that k} (i) is determined, and the increment value Δu _k (i) of the correction command is obtained by Δu _k (i) = Δu _k-1 (i) + Δσ _k (i) and output to the control target. Learning control method. 2. A control system for inputting an incremental value of a target command and an incremental value of an output of a controlled object and outputting a correction command to the controlled object so that the output of the controlled object follows a target command repeating the same pattern. The correction command u _k (i) is obtained by integrating the increment value Δu _k (i) of the correction command obtained in claim 1,
The learning control method according to claim 1, wherein the learning control method outputs to a controlled object.