JPH0250701A - Adaptive control method - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an adaptive control method that compensates for changes in process characteristics.

(Prior Art) Gain scheduling type adaptive control and self-tuning type adaptive control are conventionally known as methods for compensating for process characteristic variations.

First, gain scheduling type adaptive control will be explained with reference to FIG.

In this control, the controller 12 inputs the deviation y-y between the set value y from the subtracter 11 and the feedback amount y from the process 13, and also inputs the deviation y-y from the scheduling element 14.
The control parameter K is taken in from. On the other hand, when the optimal adjustment value for a control parameter changes due to a change in the characteristics of the process 13, the scheduling element 14 measures the nonlinear factor (environmental condition) that caused the change, and uses an adaptive system that indicates the detection result. The control parameter K is updated according to the index P. Then, the controller 12 outputs the optimum operating variable U to the process 13 based on the updated control parameter K. Note that the value of the control parameter to be updated by the scheduling element 14 can be determined by a gain schedule curve (GS curve) representing the relationship between the adaptation index P and the control parameter.

This gain scheduling type adaptive control has an adaptation index P
This method has the advantage of being able to cope with sudden changes in characteristics as long as the correspondence between and the control parameters is accurately known (in other words, as long as the GS curve is determined).

Next, self-tuning type adaptive control will be explained with reference to FIG.

In this control, the controller 12 inputs the deviation y-y between the set value y from the subtractor 11 and the feedback amount y from the process 13, and also takes in the control parameter K from the parameter adjuster 15.

Then, the process characteristic estimating means 16 estimates the characteristics of the process 13 based on the fact that the characteristics have changed, without measuring the environmental conditions of the process 13, in other words, without determining the adaptation index P, Parameter adjustment means 15
is the control parameter K optimally adjusted based on this estimation result.
The controller 12 controls the process 13 using the operation variable U corresponding to the control parameter K.

This self-tuning adaptive control always
Since the control parameters are adjusted by estimating the characteristics described in No. 3, it has the advantage of being able to cope with fluctuations in process characteristics that change over time, such as changes over time.

(Problem to be Solved by the Invention) However, gain scheduling type adaptive control is basically open-loop control, and a fixed GS curve must always be used. - It has the disadvantage of not being able to adapt to changes in characteristics due to aging.

Furthermore, self-tuning adaptive control requires a considerable amount of calculation to estimate process characteristics.

Determination of control parameters takes time and has the disadvantage that it cannot respond to sudden changes in process characteristics.

The present invention was proposed in order to solve the above problems, and by constantly adjusting optimal control parameters for characteristic fluctuations due to non-linear factors in the process and temporal characteristic fluctuations such as changes over time and aging. The objective is to provide an adaptive control method that allows the controller to adapt to any environment and environmental conditions that cause variations in process characteristics.

(Means for Solving the Problems) In order to achieve the above object, the present invention includes a consistency determining means that determines the relationship between the controller and the process based on the manipulated variable that is the output of the controller and the output of the process. The consistency is sequentially determined, and a consistency coefficient determined by the degree of consistency between the two is calculated, and the scheduling element calculates a gain according to an adaptation index, which is a parameter indicating the environmental conditions of the process, based on a gain schedule curve. The controller outputs to the process a manipulated variable whose control parameter is the multiplication value of the gain and the matching coefficient, while the database sequentially stores the matching coefficient and the adaptation index as paired data. Ivy and gain schedule curve update means.

When the number of paired data stored in the database reaches an arbitrary number, the gain schedule curve is updated based on each of the stored paired data, and then the number of stored paired data reaches a predetermined set number. In this case, each time new paired data is stocked, the database deletes the earliest paired data among the paired data already stocked in the database, and the gain schedule curve updating means The gain schedule curve is sequentially updated based on all pair data stored in the database for each stock.

(Operation) In the present invention, the process is feedback-controlled by the controller.

On the other hand, the consistency determining means inputs the manipulated variable that is the output of the controller and the output of the process, and calculates a consistency coefficient at regular intervals, for example. Further, a gain corresponding to the adaptation index is determined by the scheduling element based on the gain schedule curve.

Then, a control parameter of the controller is determined by a multiplication value of the gain and the matching coefficient.

A controller controls the process.

Further, the database sequentially stores paired data of the matching coefficient and the adaptation index, and when the stored paired data reaches an arbitrary number.

All the stored paired data are analyzed by an operator or the like, and the gain schedule curve updating means updates the gain schedule curve according to the analysis results.

In the database, when the number of paired data to be stored exceeds a predetermined setting value, the earliest data is deleted each time new paired data is stocked, so the database always contains a new set of new paired data. Data on the number of items is stored. Then, the gain schedule curve updating means is
Based on the latest paired data group, each time new data is stocked, a new gain schedule curve is updated that always takes into consideration changes over time and aging of the process.

(Embodiment) Hereinafter, one embodiment of the present invention will be described with reference to the block diagram of FIG. 1 along with the drawings.

Subtractor 1 is the set value y and the process variable (process output)
The deviation y from y is calculated, and this deviation is input to the controller 2.

The controller 2 outputs a manipulated variable U based on the deviation y--y and the control parameters taken from the control parameter generator 8 to the process 3 and the consistency determining means 4.

The process 3 is controlled based on this manipulated variable U, and the output y of this process 3 is fed back to the subtracter 1 and also output to the consistency determining means 4.

The consistency determining means 4, which receives the manipulated variable U and the process variable y, measures the consistency between the controller 2 and the process 3 from both variables U and y, and controls the consistency coefficient α according to the measurement result. Parameter generator 8 and database 5
Output to. here.

The control parameter generator 8 may be incorporated inside the controller 2, or may be provided outside the controller 2 as shown in FIG.

In addition to the consistency coefficient α, the database 5 includes the process 3
A parameter P (hereinafter referred to as "r adaptation index") indicating the environmental condition of the r is also input. These α and P are stored in the database 5 as paired data D.

A gain schedule curve updating means (hereinafter referred to as "rGS curve updating means") 6 calculates an optimal gain schedule curve (rGS curve updating means) from each pair of data D stored in the database 5.
g (hereinafter referred to as "JGS curve") is determined and output to the scheduling element 7 to update the GS curve.

The scheduling element 7 outputs this updated GS curve g to the control parameter generator 8.

The control parameter generator 8 calculates αXg from the GS curve g and the GS curve parameter α and outputs it to the controller 2 as a control parameter.

The above-described embodiments of the present invention will be described in more detail below.

For example, assume that the GS curve g is initially set to go shown in FIG. 2, and that the consistency determining means 4 judges the consistency between the controller 2 and the process 3 at regular time intervals. shall be.

First, the consistency determining means 4 uses the controller 2 and the process 3.
If it is determined that the control parameter K is too large than the appropriate value, set α to 1, and if it is determined that it is too small, set α>1.

When it is determined that matching is achieved, α=1 is output.

If the first detected value of the matching coefficient α is α, then the output of the control parameter generator 8 is α,×g1.

Here, in the database 5, α and the adaptation index P1 at the time of first detection are stocked as paired data D□.

Next, it is assumed that the matching coefficient α is detected an arbitrary number of times (i times), and i pieces of paired data D1 to Di are stored in the database 5. Here, the operator or the like determines whether or not the GS curve should be updated, and when it is determined that the GS curve should be updated, the GS curve updating means 6 updates the GS curve from gl to gl. The GS curve related to this update is determined by excluding abnormal paired data caused by hunting or the like. The change from gl to gl is determined, for example, by visual observation by an operator or the like based on the distribution of each pair of data D1 to Di indicated by x marks in FIG. Note that Dh in FIG. 2 is abnormal data due to hunting or the like. At this time, the control parameter generator 8 sets α1×g2 as a control parameter for the controller 2 (see FIG. 2).

Then, it is assumed that the matching coefficient α is sequentially detected and the number of pairs of data stored in the database 5 reaches the set number N. Here, the operator or the like updates the GS curve from gl to g3 using the GS curve updating means 6. The GS curve related to this update is determined based on the N pieces of paired data Dyu to DN, excluding abnormal paired data due to hunting etc., as in the update from gl to gz described above. The value of the control parameter at this time is αsXgi. Note that the set number N is determined depending on the characteristics of the control system.

Next, the consistency determining means 4 performs the next detection, and when the consistency determining means 4 outputs αN・1, the database 5 erases the first paired data D1, and the second to N+1th paired data D ,, D3,..., DNN13 is stocked. Then, the OS curve updating means 6 uses these paired data D.
Based on z ”' D N + x, GS
Update the curve from g to g. Here, the output of the control parameter generator 8 is αN◆z X g 4.

Similarly, the GS curve is sequentially updated to gi*gs*..., and the controller 2 performs adaptive control on the process 3 in response to changes over time and aging.

FIG. 3 is a graph showing how the matching coefficient α changes at this time, with time t plotted on the horizontal axis.

Here, t-t shows the case where α is set small because the control parameter K is detected to be too large, and t2 shows the case where α is set large because the control parameter K is detected to be too small.

In the above embodiment, the GS curve is updated once until the number of pairs of data D stored in the database 5 reaches the predetermined set number N, but this may be done multiple times, or it may be Depending on the nature of the system, the set number N
You don't have to repeat the steps until you reach .

In addition, in the above embodiment, from g of the GS curve to gl, g
The output K of the control parameter generator 8 when updating, updating, etc.
αNX g3. αN*i
It may also be Xg3, . . .

Furthermore, in the above embodiment, when the number of pairs of data D stored in the database 5 becomes N or more, the GS curve g is updated every time a new pair of data D is stocked. The GS curve g may be updated every cycle (for example, every two stocks).

Note that the present invention can be applied to multivariable control.

In this case, set value Y” + feedback it y
+ manipulated variable U, controller output U, adaptation index P.

The matching coefficient α, the paired data D, the control parameters, and the GS curve are treated as vector variables.

(Effects of the Invention) The present invention relates to data related to the consistency between the controller and the output of the process <g1 summation coefficient), and constantly measured values of the environmental conditions of the process that change over time from the past to the present. Since we decided to update the gain schedule curve sequentially based on a large number of pairs of data (adaptation index), it is said that it cannot adapt to slow characteristic changes over time, which is a drawback of gain scheduling type adaptive control. The disadvantages can be eliminated, and the disadvantage of self-tuning adaptive control, which is that it cannot respond to sudden changes in characteristics, can be eliminated.

That is, according to the present invention, it is possible to always adjust optimal control parameters for characteristic fluctuations due to nonlinear factors of process characteristics and temporal characteristic fluctuations such as changes over time and aging, resulting in fluctuations in process characteristics. It becomes possible to provide an adaptive control method that adapts the controller to all environmental conditions.

[Brief explanation of the drawing]

FIG. 1 is a block diagram for explaining an embodiment of the adaptive control method according to the present invention, FIG. 2 is a graph showing a gain schedule curve in the embodiment shown in FIG. 1, and FIG. 3 is a graph of the matching coefficient. Graph showing changes over time, 4th
The figure is a block diagram for explaining a gain scheduling type adaptive control method as a prior art, and FIG. 5 is a block diagram for explaining a self-tuning type adaptive control method. 1... Subtractor, 2... Controller, 3...
...process, 4...consistency determination means. 5...Database, 6...OS curve update means,
7... Scheduling element, 8... Control parameter generator

Claims (1)

[Scope of Claims] The consistency determining means sequentially determines the consistency between the controller and the process based on the manipulated variable that is the output of the controller and the output of the process, and determines the consistency coefficient determined by the degree of consistency between the two. and the scheduling element is
A gain corresponding to an adaptation index, which is a parameter indicating the environmental condition of the process, is calculated by a scheduling element based on a gain schedule curve, and a manipulated variable whose control parameter is a multiplication value of the gain and the matching coefficient is calculated by the controller. On the other hand, the database sequentially stocks the matching coefficient and the adaptation index as paired data, and when the paired data stored in the database reaches an arbitrary number, the gain schedule curve updating means The gain schedule curve is updated based on each of the stored paired data, and thereafter, when the number of stored paired data reaches a predetermined set number, the database is updated every time new paired data is stocked. In addition, the gain schedule curve updating means deletes the earliest pair data among the pair data already stocked in the database, and the gain schedule curve updating means updates the gain schedule curve based on all the pair data stored in the database for each new stock. An adaptive control method characterized by sequentially updating a gain schedule curve.