JPH0363704A - Model norm type adaptive controller - Google Patents

Abstract

PURPOSE:To settle a problem of the model follow-up deterioration when there is a zero point by a simple constitution, and also, to adjust the model follow-up gain by providing a filter having the same pole as a zero point of a plant. CONSTITUTION:A filter 2 having a pole of a value being equal to a zero point of a plant 1 is inserted into this side of an input of the plant 1, and the zero point of the plant 1 is offset. Also, a means 7 for managing a follow-up state to a model 4 is provided, and the gain of an adaptive system is adjusted. In such a way, when the filter 2 having a pole of the same value as the zero point is inserted into the input of the plant 1, the zero point of the plant 1 is offset, and it becomes equivalent to the plant having no zero point. Also, when the model follow-up performance is insufficient, the gain is raised by the managing means 7, and as a result of raising the gain, when the model follow-up performance is deteriorated, the adjustment to correct gain is executed by lowering the gain by the managing means 7.

Description

[Detailed description of the invention] Industrial application field of the present invention (Part 1) 8. Adaptive controller that can control even if the characteristics of the controlled object (plant) change.
Although it concerns an adaptive controller in which the target follows a reference model, a model-based adaptive controller has been proposed to perform adaptive control on a control target that includes parameters that are unknown to conventional technology. There are two types of controllers: one has the function of varying parameters as an adaptive mechanism, and the other uses model matching to follow the model without changing parameters.An example of an adaptive controller with variable parameters is
The structure of the plant, which is described in detail in Kunihiko Ichikawa's ``Adaptive Control'' (Shokoi 1984), must be known, and furthermore, the plant parameter fluctuations must be relatively slow. Model matching does not require a complete knowledge of the plant structure, and is effective even for plants where plant parameters change relatively quickly. The diagram shows the configuration of a conventional adaptive controller based on the transfer function P (s) of the plant 11 and the transfer function Pm (s) of the model 14.
, the transfer function H(s) of the compensation element 13 are connected as shown in FIG. 3(a). In other words, the target value r is input to the model 14 and is also input to the comparator 15. The output y of the plant 11 is compared with the model plant output 7m in the comparator 16, and the result is sent to the compensation element 13 manually. ) The configuration of the compensation element 13 will be explained next. shall be.

In other words, the plate H(s) = K/P m(s). At this time, the transfer characteristic (y /
r = (1+K) ・P(s)/ (−P(S to 1,000)
)/Pm(s)) Then, when K→■, the value of y/r(yo P'm(S)
If we increase Tonan Sunawa - &K, the characteristics of the system will approach the characteristics P m (s) of model 14. In other words, an input/output relationship unrelated to the characteristic P(s) of the plate plant 11 is obtained.

Problems to be Solved by the Invention In the conventional example explained above (although some shortcomings have been pointed out) number 106
8 (Oct, , 25, ° 88) (top) points out that when a plant has a zero point, errors occur in the plant's response to the model.

Theoretically, the ideal gain for the compensation element is the force t, which is infinite.In actual control systems, it may not be desirable to make the value of K too large due to noise, etc., and it is difficult to provide an appropriate gain. A method to solve the problem even if necessary is to insert a filter with a pole equal to the plant zero point before the plant input, cancel the plant zero point, and provide a means to manage the tracking state of the model. , adjust the gain of the adaptive system.

If a filter with poles of the same value as the zero points is inserted into the input of the working plant, the zero points of the plant will be canceled and the plant will become equivalent to a plant without zero points.In addition, if the model tracking performance is insufficient, the control means will increase the gain. If the viscosity model tracking performance deteriorates when the gain is increased, the gain can be adjusted to an appropriate level by lowering the gain using the management means. In the block diagram showing the configuration of the standard adaptive controller, the target value r is manually input to the comparator 5 and the model plant 4, and the output ym of the model plant 4 is sent to the comparator 6. In the comparator 5, the compensation element 3 The output of the filter 2 is sent to the plant 1, the output y of the plant 1 is sent to the comparator 6, and the output y of the model plant 4 is sent to the comparator 6. The results are compared and inputted into the compensation element 3.The management means 7 examines the state of the control system and adjusts the parameters of the control system, and the adjustment method will be described later.The management means 7 is configured by, for example, a computer. Next, we will explain the configuration of compensation element 3. Compensation element 3 has a gain of 1<
A sword composed of things multiplied by + Generally model plant 4
The inverse characteristic of Pm(s) is that the order of the numerator is higher than the order of the denominator, and it is difficult to construct one with good stability.
In Fig. 1, Ps(s) is added to prevent the order of the numerator from being higher than the denominator.Next, we will explain filter 2.When plant l has a zero point, plant 1 The characteristic P (s) can be written as follows: P (s) = P '(s) · (1 + Ts) In this case, filter 2 has the same pole as the zero point of plant l (s = 1/T). Design as follows.

Also, if the gain of plant 1 is insufficient and the model tracking is not good, you can easily determine whether the model tracking is good even if you give a gain of 2 to filter 2 (by looking at the output signal of comparator 6). However, Figure 2 is a flowchart showing a gain adjustment method.The means for executing this flowchart is the control means 7 in Figure 1.The execution result is sent to the compensation element 3 or filter 2 in Figure 1. .

The processing shown in this flowchart is repeated at regular intervals. First, in process 20, the absolute amount of error between the model plant output ym and the actual plant output y is determined.Next, the process proceeds to judgment 21, where it is checked whether the absolute amount of error is greater than a certain threshold value.Absolute amount of error If is larger than the threshold, proceed to judgment 22. If it is smaller, proceed to process 25. On the other hand, in judgment 22 (above), compare it with the absolute amount of error from the previous check, and if it has increased, proceed to process 23. If it has decreased, proceed to process 24. In process 23, change the direction of gain correction. Nawasaka: If the gain has been increasing so far, decrease the gain, and if the gain has been decreasing, increase the gain. This means that there was an effect, so we will continue to modify the gain.

By repeating the above process, an appropriate tracking gain can be obtained.Also, when there is a lot of noise and the absolute amount of error does not become less than the threshold value, L gain adjustment will minimize the absolute amount of error. Since the fine adjustment is repeated in each condition, it is difficult to obtain an appropriate gain. Of course, it can be applied.

As described in detail, the present invention has a simple structure and is transferable.
It is possible to provide a model-based adaptive controller that can solve the conventional problem of model tracking deterioration when there is a zero point, and can also adjust the model tracking gain.

[Brief explanation of drawings]

Claims (2)

[Claims] (1) Set a reference model for the plant, input the difference between the response of the reference model and the response of the plant into a compensation element based on the inverse characteristics of the reference model, and calculate the difference between the output of the compensation element and the target value. is a model-based adaptive controller that takes as an input to the plant, the difference between the output of the compensation element and the target value is input to a filter having the same pole as the zero point of the plant, and the filter output is used as the plant input. A model-based adaptive controller characterized by the following configuration: (2) Set a reference model for the plant, input the difference between the response of the reference model and the response of the plant into a compensation element based on the inverse characteristic of the reference model, and calculate the difference between the output of the compensation element and the target value. is a model-based adaptive controller that takes as an input to a plant, the absolute amount of the difference between the model response and the plant response is checked, and if the difference exceeds a threshold, the compensation element or the plant input information is A model-based adaptive controller characterized by comprising: a management means for changing the gain of the controller and reversing the direction of the gain change if the absolute amount of the error increases as a result of the change.