JPS62219007A - Identification arithmetic unit of control parameter - Google Patents

Abstract

PURPOSE:To obtain an identification arithmetic unit of a control parameter where the influence of noise is reduced and the identification precision is improved, by weighting the control parameter with a weight coefficient to identify the control parameter. CONSTITUTION:A control system consists of a control system 1 and a device 2 to be controlled, and an identification arithmetic unit 7 to which an input value U(k) and an output value Y(k) of the control system 1 are inputted to obtain an arithmetic output is provided. The input value U(k) and the output value Y(k) of the control system 1 as the identification object are inputted to an identification computing element 3 to obtain an operation output 4. The input value U(k) and the output value Y(k) are inputted to an excitation condition decider 5 also, and the excitation condition decider 5 decides an excitation condition on a basis of these values. A decided level signal from the excitation condition decider 5 is inputted to a weight coefficient generator 6, and a weight coefficient (m) is calculated on a basis of this signal. The weight coefficient (m) is inputted to the identification computing element 3, and the control parameter which controls the control system 1 is weighted to obtain the arithmetic output 4.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Field of Application) The present invention relates to a control parameter identification calculation device that can identify control parameters of a model formula defining characteristics of a control system with high accuracy.

(Prior art) Control parameters are identified in order to understand the characteristics of control systems and detect abnormalities.
No. 9-014194 has already been filed by the applicant of the present invention.

In order to improve the accuracy of this identification calculation, it is necessary to have a sufficient number of data so that the value of the identified control parameter converges, and an identification excitation input that can obtain the required data.

However, the normal identification excitation input does not always satisfy the above-mentioned conditions for increasing the accuracy of identification.

In a conventional control system in which stable control is performed, the timing at which the identified excitation input causes a large fluctuation that satisfies the identification condition is limited.

Furthermore, the duration of fluctuations is often short.

(Problems to be Solved by the Invention) Therefore, if sufficient identification conditions are not provided, the identification accuracy will not be sufficient under normal control conditions. If input fluctuations exceeding a predetermined value are forcibly applied to improve this problem, there is a problem in that excessive stress is applied to the controlled equipment.

SUMMARY OF THE INVENTION Therefore, it is an object of the present invention to provide a control parameter identification calculation device with time-varying weights that can analyze input and output signals of a control system to be identified and identify them with high accuracy.

[Structure of the invention]

(Means for Solving the Problems) A control parameter identification calculation device according to the present invention uses a control parameter identification calculation device to calculate a control parameter based on an input data string to a control system that controls a controlled device and an output data string from the control system. An identification calculation device that identifies control parameters of an identification calculation formula includes an excitation condition determiner that inputs an input data string and an output data string, determines an excitation condition, and outputs a determination level according to the excitation condition; The present invention is characterized in that it includes a weighting factor generator that inputs a level and generates a weighting factor according to the determination level, and identifies the control parameter by weighting the control parameter with the weighting factor.

(Function) The weighting coefficient calculated by the identification calculation device according to the present invention is introduced into the calculation formula in the identification calculation unit, and the weight is increased for data where the identification excitation condition is good, and conversely, when the identification excitation input is small, weighting is applied. A highly accurate identification calculation output is obtained by decreasing . It is assumed that the identified excitation condition is determined to be good when the following conditions are satisfied from the input/output data string.

■ When the signal level is large in the input data string.

■ When the signal level difference between adjacent identification steps is large in the output data string.

■ Considering the delay between input and output, identify the interval for the output delay after the condition of ■ is satisfied.

■ When it is determined that the signal-to-noise ratio is large from frequency analysis of the input data string.

■ When input/output data strings and parameter predicted values are entered into a model formula to obtain a residual term, and the residual term becomes less than the threshold value.

■ When the conditions from ■ to ■ above are arbitrarily combined.

■ When the operating conditions of the system allow it to be determined that there are few disturbances.

(Example) Examples of the present invention will be described in detail below with reference to the accompanying drawings.

FIG. 1 shows a system configuration diagram of an identification calculation device according to the present invention and a control system to which the device is applied.

The control system is composed of a control system 1 and a controlled device 2, and the input value U (k) and the output value Y (k
) is provided to obtain an arithmetic output. The input value U (k
) and the output value Y (k) are input to the identification calculator 3, and a calculation output 4 is obtained.

The input value U (k) and the output value Y (k) are also input to the excitation condition determiner 5. The excitation condition determiner 5 determines the excitation condition based on these values.

The determination level signal ρ from the excitation condition determiner 5 is input to the weighting coefficient generator 6, and the weighting coefficient m is calculated based on this.

The weighting coefficient m from the weighting coefficient generator 6 is input to the identification calculator 3, which weights the control parameters defining the control system 1 to obtain a calculation output 4.

Below, the input value U (k) is the angular velocity signal θ (k) of the rotating body,
The present invention will be described by taking as an example a control system in which the control system 1 is an integral element, the output value Y (k) is an angle signal θ (k) of a rotating body, and the controlled device 2 is a rotating device whose angle is controlled. .

Assuming that the sampling step for the control system is 1 and the system noise is v (k), the state equation is expressed by the following equation.

Here, it is assumed that the control parameters [a, bl in equation (1) are identified by the identification calculation device 7.

FIG. 2 is a waveform diagram showing input and output signals in the control system 1. Where the absolute value of the angular velocity θ(k) is small and close to O, the system noise v (k) is larger than the angular velocity θ(k), and an oscillating waveform is observed.

Therefore, by increasing the weighting of the portion where the angular velocity θ(k) is large, the influence of noise can be reduced, and it can be expected that the identification accuracy will be improved.

FIG. 3 is a flowchart showing the flow of excitation condition determination in the excitation condition determination unit. As shown in FIG. 3, it is determined whether the absolute value of the angular velocity θ(k) is larger than the threshold value η (step 51), and whether the signal of the angle θ(k) has increased or decreased (step 52). ) to set the determination level signal p from ρ1 to fJ3.

That is, when the absolute value of the angular velocity θ(k) is greater than the threshold η and the angle θ(k) is increasing, the determination level signal 1 is output as 13 (step 53).

If the absolute value of the angular velocity θ(k) is greater than the threshold η and the angle θ(k) is not increasing, the determination level signal 1 is output as 12 (step 54).

Furthermore, if the absolute value of the angular velocity θ(k) is not greater than the threshold η, the judgment level signal ρ is set! Output as J1.

FIG. 4 is a flowchart for explaining the operation of the weighting coefficient generator 6 shown in FIG. Judgment level signal 1 is 1
or j2 (steps 61 and 62)
, a weighting coefficient m is generated according to the value. That is, when the judgment level signal 1=113, the weighting coefficient becomes m-3 (step 63), when the judgment level signal ρ=12, the weighting coefficient m=2 (step 62), and in other cases, the weighting coefficient m=1. The following operations are performed (step 65).

In the case of input/output signals as shown in Figure 2, m=1 in sections t1 and t3, m÷3 in section t2, and m=1 in section t4.
becomes. That is, when the angle increases in the interval t2, the excitation input is large and the excitation conditions are the best, so the weighting coefficient g is increased, and in the interval t1 or t3, the l1NJ vibration condition is poor, so m is decreased.

Next, an example of a method for introducing the weighting coefficient m into the identification calculation formula in the identification calculation unit will be explained. Assuming that the least squares method is applied to the identification calculation, [a, bl...(2).

That is, for each sampling step, the input/output data is used a plurality of times as many times as the weighting coefficient m at that time to obtain the weighting of the control parameters.

FIG. 5 is a characteristic diagram illustrating the effect of introducing weighting coefficients. (
a, b) On the plane, (a, bo) are the true identified values.

If data from section t1 with poor excitation conditions is used, a value (a, bl) far away from the true identified value (a, bo) will be identified; if data from section t2 where excitation conditions are good is used, the value (a, bl) will be identified.
ao.

The values (a, b2) close to b ) are identified. If a large amount of data in the interval t2 is used across the range 0 and 0, the value (a□
, b□).

FIG. 6 is a block diagram showing another embodiment of the present invention.

Although the schematic configuration is almost the same as that shown in FIG. 1, the configuration of the excitation condition determiner 5 and the configuration of the weighting coefficient generator 6 are different.

The excitation condition determiner 5' according to the present invention has an input data string U (
k), perform frequency analysis often used in signal processing to calculate the signal-to-noise ratio (S/N) ratio, and calculate the S/N ratio.
A weighting coefficient generator 6 from a judgment level signal 1 proportional to the N ratio
', a weighting coefficient m is generated and sent to the identification calculation 33 at every predetermined step by a switch operated by a signal k output from the fixed operator 3. The weighted least squares method is adopted as the identification calculation formula in the identification operator 33, and the weighting coefficient m is used as a coefficient that changes the weighting coefficient matrix in a time-varying manner.

That is, the identification calculation formula is given by the following formula.

[a, bl Here, m=weighting coefficient W=simple matrix, that is, the identification calculation formula is given by equation (3).

In this configuration, the identification excitation condition is determined by the S/N ratio, so even if the input signal width appears to be large in the amplitude judgment of the input signal due to noise, if the S/N ratio is small, weighting is not performed. Conversely, even when the fixed excitation input appears to be small, if the S/N ratio is large, the weighting can be increased.

Also, in the identification calculation, the weighting coefficient m allows finer adjustment than when using the normal symmetric square method.

〔Effect of the invention〕

As explained above based on the embodiments, in the present invention, the input/output signals of the control system to be identified are analyzed in the normal control state, and when the noise level is relatively low, weighting is increased in the identification calculation. Conversely, when the noise level is relatively high, the weighting is reduced.

Therefore, it is possible to obtain a control parameter identification performance device in which the influence of noise is reduced and identification accuracy is improved.

[Brief explanation of drawings]

FIG. 1 is a diagram of a control system to which the present invention is applied;
Figure 2 is a waveform diagram showing an example of input/output signals, and Figure 3 is a waveform diagram showing an example of input/output signals.
FIG. 4 is a flowchart for explaining the operation of the excitation condition determiner in FIG. 1, FIG. 4 is a flowchart for explaining the operation of the weighting factor generator in FIG. 6 are block diagrams showing other embodiments of the present invention. 1... Control system, 2... 1 controlled device, 3... Identification calculator, 5.5'... Excitation condition determiner, 6.6'...
・Weighting coefficient generator, ρ...judgment level signal, m...
weighting factor. Applicant's agent: Mr. Sato! Figure 4 Figure 5 Figure 6

Claims (1)

[Scope of Claims] 1. Identification calculation for identifying control parameters of an identification calculation formula for the control system based on an input data sequence to a control system that controls a controlled device and an output data sequence from the control system inputting the input data string and the output data string in the device;
an excitation condition determiner that determines an excitation condition and outputs a determination level according to the excitation condition; and a weighting factor generator that inputs the determination level and generates a weighting coefficient according to the determination level, A control parameter identification calculation device, characterized in that the control parameter is identified by weighting the control parameter with a coefficient. 2. The excitation condition determiner outputs a determination level for generating a small weighting coefficient for an input value of small amplitude, and a determination level for generating a large weighting coefficient for an input value of large amplitude. A control parameter identification calculation device according to claim 1. 3. The control parameter identification calculation device according to claim 1, wherein the excitation condition determiner determines the S/N ratio by frequency analysis and outputs a determination level according to this S/N ratio.