JPH02308303A - Automatic adjustment controller - Google Patents

Abstract

PURPOSE:To automatically control a control parameter so that the command value response and the disturbance response are satisfactory together by providing a disturbance response simulating means and a parameter inferring mechanism which simultaneously evaluates the command value response and the disturbance response to determine the parameter of a controller. CONSTITUTION:A command value (r) stored in a storage device 5, an output yd of a disturbance response computing element 2 which simulates the disturbance in accordance with the manipulated variable from a control arithmetic unit 1 having the same characteristic as a controller 3, and an output (y) of a control object 4 are taken into a feature quantity extracting means 6 to generate a feature quantity, and it is inputted to a parameter inferring mechanism 7 to determine the parameter of the controller 3. Consequently, parameters of the control arithmetic unit 1, the controller 3, and the disturbance response computing element 2 are controlled by a parameter control means 8 in accordance with the output of the parameter inferring mechanism 7. Thus, control characteristics of good command value response and disturbance response are obtained in the controller 3.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention (Industrial Application Field) The present invention relates to an automatic adjustment control device used for controlling an electric motor, for example.

(Prior Art) For example, when initially adjusting a control device that controls an electric motor, the key point of the adjustment is to make both the target value response and the disturbance response favorable.

However, the target value response and the disturbance response are contradictory, and it is necessary to balance them in order to adjust these responses. For this reason, in the past, a skilled person checked the balance between the target value response and the disturbance response while checking the response waveform using an oscilloscope, etc., and made adjustments.

(Problems to be Solved by the Invention) However, such an initial adjustment method for a control device has a problem in that it requires a lot of effort and time even for an expert.

Recently, there has been a desire to automate such adjustments, but the reality is that there is currently no means to realize such a desire.

The present invention provides an automatic adjustment control device that automates the initial adjustment of the control device and can respond to changes in characteristics of the controlled object such as parameter changes by intermittently disabling the zero-motion adjustment mechanism during operation. The purpose is to share ideas.

[Structure of the Invention] (Means for Solving the Problem) In order to achieve the above object, the present invention provides an automatic adjustment control device having the following structure.

FIG. 1 is a functional block diagram for explaining the present invention. In FIG. 1, the control device 3 inputs the command value r and the output y of the controlled object 4 to control the controlled object 4, and the control calculation device 1 has the same parameters as the control device 3. , the command value r, and the output y fed from the disturbance response computing unit 2 are inputted to generate a manipulated variable for the disturbance response computing unit 2 that simulates the disturbance response. The storage device 5 stores the output y of the disturbance response calculator 2, the output y of the controlled object 4, and the command value r, and the feature amount extraction means 6 generates the feature amount of the controlled object from the data in the storage device 5. Input to parameter inference mechanism 7. The parameter inference mechanism 7 determines the parameters of the control device 1 from the feature quantities generated by the feature quantity extraction means 6, and inputs the parameters determined here to the parameter adjustment means 8. This parameter adjustment means 8 adjusts the parameters of the control calculation device 1, the control device 3, and the disturbance response calculation device 2 according to the output of the parameter inference mechanism 7.

(Function) In the automatic adjustment control device having such a configuration, a disturbance response is generated according to the command value r stored in the storage device 5 and the operation amount from the control calculation device 1 having the same characteristics as the control device 3. The output yd of the disturbance response calculator 2 to be simulated and the output y of the controlled object 4 are taken into the feature quantity extraction means 6 to generate a feature quantity,
By inputting this to the parameter inference mechanism 7, the parameters of the control device 3 are determined. According to the output of the parameter inference mechanism 7, the parameter adjusting means 8 controls the control arithmetic device 1, the control device 3, and the disturbance response arithmetic unit 2. It becomes possible to adjust the parameters of the control device 3, so that both the command value response and the disturbance response of the control device 3 can be made to have good control characteristics.

(Example) An embodiment of the present invention will be described below with reference to the drawings.

FIG. 2 shows an embodiment of an automatic adjustment control device using a fuzzy inference mechanism as the inference mechanism.

In FIG. 2, the control device 3 has three parameters: an integral gain Ki1, a first proportional gain Kp5, and a second proportional gain Ka. The integral gain Ki has the effect of reducing the steady-state deviation to zero, the first proportional gain Kp is an amount that determines the speed of rise of the response, and the second proportional gain Ka has the effect of suppressing the amount of overshoot. Further, S in the block of the control device 3 is a Laplace operator, and 1/S represents an integral. The control calculation device 1 is for disturbance response, and has exactly the same structure as the control device 3, and has the same parameters. It is assumed that the controlled object 4 controls an electric motor. The disturbance response calculator 2 is a part 21 in which the converter is approximated by a first-order lag model.
, Part 2 expressing the inertia of the motor using Jd as a parameter
2. It consists of a disturbance generator 23 that generates a step-like disturbance Td, and sends out an output ωd. Storage device 5 is controlled object 4
The output ω of the disturbance response calculator 2, the output ωd5 of the command value ω", and the momentary values of the command value ω" are stored.The time responses of ω and ωd indicate the target value response and the disturbance response, respectively.The feature extraction means 6 The feature quantities to be extracted are selected based on the command value ωr, and each feature quantity is calculated.For example, when the command value ωr changes stepwise, the feature quantities include the rise time, overshoot amount, delay time, and The impact drop amount and settling time of the disturbance response are selected.The parameter inference mechanism 7 includes a fuzzy inference execution unit 71 and a fuzzy rule base 7.
Consists of 2.

Here, the fuzzy rule base 72 is a rule base written using seven types of ambiguous expressions (for example, quite small, small, somewhat small, appropriate, somewhat large, large, and quite large).

These seven types of expressions are defined by functions as shown in FIG.

Note that the horizontal axis indicates the normalized feature amount.

By mapping with the function shown in FIG. 3, each feature quantity is converted into a numerical value of O to 1. This function is called the membership function in fuzzy inference. The definition of the membership function shown here is an example, and the shape of the function is defined for each parameter.

The fuzzy inference execution unit 71 uses the fuzzy rule base 71 to execute the following inference.

(1) Evaluate each rule and obtain inference results for each rule.

The condition part of each rule is evaluated from the feature amount obtained by the feature amount extracting means 6, and the degree of suitability (an index representing the extent to which the condition is satisfied) of each rule is determined. Next, the membership function corresponding to the conclusion part of each rule is transformed according to the degree of suitability of each rule, and an inference result is obtained.

As an example, a method of reasoning regarding the rule ``If the rise time is large and the overshoot amount is quite small, the proportional gain Kp should be made quite large'' will be explained. FIG. 4 illustrates this process. In this example, the suitability of the conditions "rise time is large" and "overshoot amount is quite small" is 0.7 and 0.3, respectively. In this rule, two conditions are connected by "and (inference part)", so the minimum value of each fitness is taken and 0.3 is set as the fitness of the condition part. ``Make the proportional gain Kp considerably larger'' in the inference section is also expressed by the membership function shown in FIG. 3, similar to the condition section, and this function is modified depending on the degree of conformity of the condition section.

In this case, the fitness of the condition part is 0.3, so the membership function that says "Kp is considerably large" has a maximum value of 0.
．． Transform it so that it becomes 3.

(2) Determine parameters by integrating the inference results of each rule.

The membership functions obtained as the evaluation of each rule are combined for each parameter to determine each parameter. Figure 5 illustrates the decision process. The overall procedure is
First, find the minimum value of all membership functions for one parameter, and use this as the membership function for that value (this corresponds to taking the logical sum of the inference results of each rule). Next, find the centroid of the membership function of each parameter and use this as the inferred value of the parameter.

Here, the center of gravity Gp of the membership function A (x) is calculated by the following equation.

Gp=J'A (x)xdx/, J'A (x)dx When the parameter is determined by the inference mechanism 7, the parameter adjustment means 8 adjusts the parameter Kp of the control device 3 and the control calculation device 1 according to the parameter. ,Ki,Ka
and a parameter Jd representing the inertia of the electric motor.

As described above, this embodiment allows ambiguous expressions such as "slightly large" and "fairly small" in the rules of the inference mechanism, and therefore has the feature that knowledge of an expert can be easily converted into rules. Furthermore, since the inference result is obtained by only one stage of inference, the inference time remains constant regardless of the value of the feature amount.

In addition to this embodiment, the following method may be used as an inference mechanism using fuzzy inference.

That is, when determining the inference result for each rule, the membership function of the conclusion part is weighted by the degree of suitability of the condition part, and the result is determined as the inference result.

FIG. 6 illustrates the changes in the membership function in this case.

Further, in this embodiment, maximum value calculation and minimum value calculation are mainly used as fuzzy inference calculations, but various calculations have been proposed for fuzzy inference, and any of these calculations may be employed.

Next, as another embodiment of the present invention, a case will be described in which a parameter inference mechanism using a calculation mechanism as shown in FIG. 7 is used. As shown in FIG. 7, this parameter inference mechanism includes an input layer 10, an intermediate layer 20, and so on. It consists of an output layer 30, as shown in FIG.
The input layer as shown in a) and (b) has a single-manpower unit with one output, and the intermediate layer and output layer have a large-power unit with one output as a constituent unit. Input layer units 101 to 105
consists of the same number of units as the feature quantity sent from the feature quantity extracting means, and inputs the feature quantity multiplied by a weight 106. The outputs multiplied by various weights serve as inputs to a plurality of units in the intermediate layer. Similarly, the outputs of the intermediate layer units 201 to 204 are multiplied by a weight 205 and then input to a plurality of output layer units. The output layer includes the same number of output layer units 301 to 304 as the parameters to be determined, and outputs the inference results for each parameter. In the example of FIG. 2, Kp+Ki.

The output layer may be composed of four units for Ka and Jd.

Here, the output of each unit is expressed by the following equation.

y-f (Σωi・yi-h) where y: output of unit, yi: output of unit i in the previous layer, ωi: weight for unit i in the previous layer, h: constant function f (x) is shown in Figure 8. The function shown in (a) is used. f
As (x), various modifications such as those shown in FIG. 8(b) can be considered.

In this method, it is important to determine the weights that connect each unit, and the feature is that the weights can be determined by a learning function by having an expert provide information on how to determine parameters for several feature quantity patterns. Several methods have been proposed as learning algorithms.

Another example is a method of using an expert system type knowledge base as the inference mechanism.

In this case, geomembrane inference can be used to reach a conclusion by repeatedly adapting the rules, so this method is advantageous for controlled objects that require complex reasoning or reasoning that uses mathematical information. becomes.

The present invention can use both initial parameter adjustment and parameter modification during operation, but in the case of initial adjustment, the original command value may not include sufficient frequency and all parameters may not be inferred. In such a case, a command value generator 13 and a switch 14 are provided internally as shown in Fig. 9, and the output of the command value generator 13 is used as the command value in the initial adjustment. This can be handled by switching to the command value.

[Effects of the Invention] As described above, according to the present invention, by providing a means for simulating a disturbance response and a parameter inference mechanism that simultaneously evaluates a command value response and a disturbance response to determine parameters of a control device, It is possible to provide an automatically adjusting control device in which control parameters are automatically adjusted so that both value response and disturbance response are good, and the precision and reliability of the control device can be improved without requiring adjustment by a skilled operator. .

[Brief explanation of the drawing]

Fig. 1 is a functional block diagram for explaining the present invention in detail, Fig. 2 is a block diagram showing an embodiment of the present invention, and Figs. 3 to 5 are for explaining the operation of the embodiment. diagram,
FIGS. 6 to 9 are diagrams for explaining other embodiments of the present invention. DESCRIPTION OF SYMBOLS 1... Control calculation device, 2... Disturbance response calculation unit, 3.
...Rameter adjustment means.

Claims (1)

[Claims] A control device that outputs a manipulated variable of a controlled object from a command value and a detected value of the controlled object includes a disturbance response calculator that simulates a disturbance response, and a disturbance response by inputting the output of this disturbance response calculator and the command value. a control calculation device having the same characteristics as the control device that outputs the operation amount of the calculation unit; a storage device that stores the command value, the detected value of the controlled object, and the disturbance response output from the disturbance response calculation unit; Feature extracting means for extracting feature quantities of a controlled object from data stored in the storage device; and parameter inference for determining parameters of the control device and disturbance response calculator from the feature quantities output from the feature extracting means. An automatic adjustment control device comprising: a parameter inference mechanism; and parameter adjustment means for adjusting parameters of the control device and the disturbance response calculator according to the output of the parameter inference mechanism.