CN102768528A - Detecting device and detecting method for control performance of multiple-input multiple-output control system - Google Patents

Abstract

The invention discloses a detecting device and a detecting method for control performance of a multiple-input multiple-output control system in the technical field of performance detection of control systems. The detecting device comprises a decoupling filter, a delay estimator, a single circuit performance detector and an index weighting unit. Modeling need not be performed to the control system, and accordingly calculation is simple, algorithm efficiency is high, and on-line calculation is facilitated.

Description

A kind of control performance pick-up unit and method of multiple-input, multiple-output control system

Technical field

The invention belongs to the Performance Detection technical field of control system, relate in particular to a kind of control performance pick-up unit and method of multiple-input, multiple-output control system.

Background technology

At industrial control field, the control loop of production run constantly increases, and production technology requires to improve day by day to the control of production run.According to statistics; There is problem that 60% controller nearly exists poor performance in current a lot of Industry Control system, and the control loop number of current production run is many simultaneously; Simple artificial examination and maintenance, time-consuming, the effort, and the reduction of control system performance of relying on; Except meeting reduce product quality, increase operating cost, reduce equipment service time, also possibly cause safety problem; On the other hand, even if the operation at the beginning of performance better controlled device, along with the prolongation of service time, controller also can wear out, control performance can correspondingly reduce.Reason for the potential degeneration of controller is many-sided; At first be exactly current disturbance and dead band; And system other dynamically as irreversible zero point etc.; What controller parameter was adjusted besides is improper, also has strain, the fault of equipment, the hardware failure of control system etc. all can become the potential cause that controller performance is degenerated.If we can detect the performance of control system effectively, will increase work efficiency greatly, reduce production costs, improve the validity of control.If can the performance of control system be detected effectively, will increase work efficiency greatly, reduce production costs, improve the validity of control.

Mimo systems is ubiquitous in the industrial process, and in reality, it is carried out performance monitoring and detect that important current demand is arranged.But because its complex structure; All there is coupling between each variable and between the loop; The factor that influences its control performance is many; Make its Performance Detection complicated, be difficult in actual industrial process, use, have important theory based on this succinct practical multiple-input, multiple-output control system Performance Detection algorithm and be worth and using value.

Summary of the invention

Complicated to mentioning the existing system Performance Detection in the above-mentioned background technology, be subject to deficiencies such as interference, the present invention proposes a kind of control performance pick-up unit and method of multiple-input, multiple-output control system.

Technical scheme of the present invention is that a kind of control performance pick-up unit of multiple-input, multiple-output control system is characterized in that this device comprises the decoupling zero wave filter, postpones estimator, single loop property detector and index weighter;

Said decoupling zero wave filter is connected with the index weighter with delay estimator, single loop property detector respectively; Postponing estimator is connected with the single loop property detector; The single loop property detector is connected with the index weighter;

The linear decoupling zero of said decoupling zero wave filter realization system output; Postpone the delay that estimator is used for calculation of filtered output; The single loop performance index of single loop property detector calculation of filtered output; The Comprehensive Control performance index that the index weighter is used to calculate the weighting coefficient of each single loop index and comprehensively obtains mimo systems.

A kind of control performance detection method is characterized in that this method may further comprise the steps:

Step 1: said decoupling zero wave filter is gathered and filtering system's output vector of multiple-input, multiple-output control system, obtains normalized vector, further obtains the eigenwert and the unit character vector of the correlation matrix of normalized vector; Obtain the filtering output matrix at last;

Step 2: the system delay that utilizes said delay estimator calculation of filtered output matrix;

Step 3: the performance index of utilizing said single loop property detector calculation of filtered output matrix;

Step 4: utilize the weighting coefficient of said index weighter calculation of filtered output matrix, the combination property detected value of controlled system.

The computing formula of said filtering output matrix is:

$\tilde{Y}=\mathrm{MY}$

Wherein:

is the filtering output matrix;

M is the matrix that the unit character vector is formed;

Y is system's output vector.

The computing formula of said performance index is:

${\mathrm{\&eta;}}_i=\frac{{\left({\mathrm{\&delta;}}_{{\tilde{Y}}_i}^2\right)}_{\mathrm{mv}}}{{\mathrm{\&delta;}}_{{\tilde{Y}}_{\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="HDA00001939833100012.png,HDA00001939833100021.png"\; state="\{\{state\}\}">i</figure-callout>}}}^2}=\frac{\underset{j=0}{\overset{d_i-1}{\mathrm{\&Sigma;}}}{f}_{\mathrm{ij}}}{\underset{j=0}{\overset{\&infin;}{\mathrm{\&Sigma;}}}{f}_{\mathrm{ij}}}$

Wherein:

η _iBe the filtering output matrix I element

Performance index;

is the filter output matrix

i-th element

The minimum variance;

is the filter output matrix

i-th element

of the variance;

d _iBe i element

The system in loop delays;

f _IjFor its noise to i element

Set up the coefficient of AR model.

The computing formula of said weighting coefficient is:

${\mathrm{\&omega;}}_i=\frac{{\mathrm{\&lambda;}}_i}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_i}$

Wherein:

ω _iBe weighting coefficient;

λ _iI eigenwert for correlation matrix.

The computing formula of said combination property detected value is:

$\mathrm{\&eta;}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{\mathrm{\&eta;}}_i$

Wherein:

η is the combination property detected value.

The present invention need not system is carried out modeling; (mutual matrix is the expansion of delay in mimo systems of singly going into singly to go out in the system more needn't to ask for the mutual matrix of system; Be not only based on object model and also calculate very complicated); Thereby calculating is succinct and efficiency of algorithm is high, is easy in line computation.Simultaneously; This invention directly uses the regular job number of control system to calculate, and need not to make an experiment, and more can not produce any influence to the normal operation of system; Be the method and apparatus of the friendly type of no invasive system, can in actual control system, use very easily.In addition, this Device Testing result is normalized, and between [0,1], it is easy to understand as a result.

Description of drawings

Fig. 1 mimo systems control performance detects figure;

The output sampling of Fig. 2 system; 1 sampling of Fig. 2 (a) system output variable; 2 samplings of Fig. 2 (b) system output variable;

The filtering output of Fig. 3 system; The filtering output variable 1 of Fig. 3 (a) system; The filtering output variable 2 of Fig. 3 (b) system.

Embodiment

Below in conjunction with accompanying drawing, preferred embodiment is elaborated.Should be emphasized that following explanation only is exemplary, rather than in order to limit scope of the present invention and application thereof.

The invention provides a kind of succinct, effectively, need priori few, be easy to calculate, actual control system normally moved multiple-input, multiple-output control system method for testing performance and the device that does not have influence.

For promoting the controlling level of multiple-input, multiple-output control system, keep watch on its performance variation, the present invention provides the method for testing performance and the device of multiple-input, multiple-output control system.As shown in Figure 1, the present invention mainly contains four parts and forms: decoupling zero wave filter, delay estimator, single loop property detector and index weighter.The each several part major function is following: the linear decoupling zero of decoupling zero wave filter realization system output; Postpone the delay that estimator calculates each filtering output; The single loop property detector is through filtering output and postpone to calculate the single loop performance index of each filtering output; The Comprehensive Control performance index that the index weighter is used to calculate the weighting coefficient of each single loop index and comprehensively obtains mimo systems.

The decoupling zero wave filter reads the close loop maneuver number to carry out outputing it to after the filtering and postpones estimator and do input; The output that postpones output and the decoupling zero wave filter of estimator is given the single loop property detector and is done input, and the index weighter of exporting to of the output of single loop property detector and wave filter is done the calculating of importing the realization integrated performance index mimo systems is carried out Performance Detection.This method and apparatus is following:

1. decoupling zero wave filter

Realize the collection and the filtering of closed loop data through the decoupling zero wave filter, the Performance Detection of mimo systems is converted into the Performance Detection of singly going into singly to go out system.The decoupling zero filtering algorithm is following:

With the output vector Y=[Y of system ₁, Y ₂... Y _n] ^TAfter normalization process was carried out in sampling, it was following to obtain normalized vector X:

X＝[X ₁,X ₂,…X _i,…X _n] ^T （1）

Wherein,

E (Y _i) be the average of the element of system's output vector;

Mean square deviation for the element of system's output vector.

The correlation matrix R of normalized vector X calculates as follows:

$R={\left(r_{\mathrm{ij}}\right)}_{n\&times;n}=\frac{1}{n}{X}^{T}X---\left(2\right)$

Obtain the eigenvalue of system through separating following equation ₁, λ ₂... λ _nWith the vectorial matrix M=[M that forms of the unit character of correspondence ₁, M ₂... M _n] ^T:

RM＝λM （3）

If R is that symmetrical matrix and eigenwert are unequal mutually, following equality is arranged:

M ^TM＝I （4）

Wherein, I is a unit matrix.

Output obtains new filtering output variable matrix

through M filtering

$\tilde{Y}=\mathrm{MY}---\left(5\right)$

Wherein, $\tilde{Y}={[{\tilde{Y}}_1,{\tilde{Y}}_2,...{\tilde{Y}}_i,...{\tilde{Y}}_n]}^T,$ ${\tilde{Y}}_i=M_i^{T}Y,$ $M_i={[m_{1i},m_{2i},...,m_{\mathrm{Ni}}]}^T.$

2. delay estimator

Usually when calculating the single loop control system performance, need to postpone as priori, so need its delay of calculating.The core algorithm that postpones estimator is following:

Corresponding to proper vector λ _i, its proper vector M _i, it is specific as follows:

M _i＝[m _1i,m _2i,…m _ni] ^T （6）

Can get by formula (5):

${\tilde{Y}}_i=m_{1i}{\mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="HDA00001939833100012.png,HDA00001939833100021.png"\; state="\{\{state\}\}">Y</figure-callout>}}_1+m_{2i}{\mathrm{<figure-callout\; id="2"\; label="Y"\; filenames="HDA00001939833100012.png,HDA00001939833100021.png"\; state="\{\{state\}\}">Y</figure-callout>}}_2+...+m_{\mathrm{ni}}{Y}_n---\left(7\right)$

Work as M _iWhen not having neutral element in the vector, can get by (7)

Delay depend on Y _i(i=1,2 ..., n), so the delay in each loop equates and is:

d _j＝min{d(Y _j)|j＝1,2,…n} （8）

Wherein, d (Y _j) expression be input to output Y _jTime lag vector.

Work as M _iThere is neutral element m _JiThe time, then by (7) Shi Kede

With Y _jIrrelevant, so

Postpone not receive Y _jInfluence.Can get in view of the above, system delay d is in the case:

Wherein,

Be M _iThe set of all neutral element rower values.

The delay in all loops has constituted system delay d.

3. single loop property detector

This single loop Performance Detection is based on the Harris index and calculates its performance at random.The single loop performance index are following usually:

${\mathrm{\&eta;}}_i=\frac{{\left({\mathrm{\&delta;}}_{{\tilde{Y}}_i}^2\right)}_{\mathrm{mv}}}{{\mathrm{\&delta;}}_{{\tilde{Y}}_{\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="HDA00001939833100012.png,HDA00001939833100021.png"\; state="\{\{state\}\}">i</figure-callout>}}}^2}=\frac{\underset{j=0}{\overset{d_i-1}{\mathrm{\&Sigma;}}}{f}_{\mathrm{ij}}}{\underset{j=0}{\overset{\&infin;}{\mathrm{\&Sigma;}}}{f}_{\mathrm{ij}}}---\left(10\right)$

Wherein:

η _iBe the filtering output matrix I element

Performance index;

is the filter output matrix

i-th element

The minimum variance;

is the filter output matrix

i-th element

of the variance;

d _iBe i element

The system in loop delays;

f _IjFor its noise to i element Set up the coefficient of AR model.

These coefficients can be set up the AR model and obtain through being carried out time series analysis.

4. index weighter

The index weighter mainly is to realize the calculating of each single loop weighting coefficient and the calculating of overall system performance.The calculating of each single loop weighting coefficient:

${\mathrm{\&omega;}}_i=\frac{{\mathrm{\&lambda;}}_i}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_i}---\left(11\right)$

Filtering is exported Its performance Index Calculation η _i, its weighting coefficient is ω _i, its comprehensive detection index is:

$\mathrm{\&eta;}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{\mathrm{\&eta;}}_i---\left(12\right)$

The method for testing performance to the multiple-input, multiple-output control system that the present invention proposes can the mode through software programming or configuration be realized in the overwhelming majority's industrial control system (device).Also can realize simultaneously through hardware.Implementation is flexible.

The controlling object transport function of certain multiple-input, multiple-output industrial process and process disturbance transport function are as follows respectively:

$T=\left[\begin{array}{cc}\frac{q^{-1}}{1-0.4q^{-1}}& \frac{K_{12}{q}^{-2}}{1-0.4q^{-1}}\\ \frac{q^{-1}}{1-0.1q^{-1}}& \frac{q^{-2}}{1-0.8q^{-1}}\end{array}\right]$ $N=\left[\begin{array}{cc}\frac{1}{1-0.5q^{-1}}& \frac{-0.6}{1-0.5q^{-1}}\\ \frac{0.5}{1-0.5q^{-1}}& \frac{1}{1-0.5q^{-1}}\end{array}\right]$

The employing controller is following:

$Q=\left[\begin{array}{cc}\frac{0.5-0.2q^{-1}}{1-0.5q^{-1}}& 0\\ 0& \frac{0.25-0.2q^{-1}}{(1-0.5q^{-1})(1+0.5q^{-1})}\end{array}\right]$

Work as K ₁₂Its Performance Detection process was following in=1 o'clock:

1, the decoupling zero wave filter is realized the collection and the filtering of system's output vector of multiple-input, multiple-output control system

The sampled data of system's two outputs this moment is respectively shown in Fig. 2 (a) and Fig. 2 (b).By formula its normalized vector X is tried to achieve with its normalization in (1), and by formula the eigenwert and the unit character vector of the correlation matrix of (2) calculating normalized vector X are respectively:

$\left[\begin{array}{c}{\mathrm{\&lambda;}}_1\\ {\mathrm{\&lambda;}}_2\end{array}\right]=\left[\begin{array}{c}1.9955\\ 0.004491\end{array}\right]$ With $M=\left[\begin{array}{cc}0.7071& -0.7071\\ 0.7071& 0.7071\end{array}\right]$

Calculate its filtering output matrix

two variable data respectively shown in Fig. 3 (a) and Fig. 3 (b) through (5).

2, postpone the delay of estimator calculation of filtered output matrix

Since there is not neutral element in the proper vector, so it is following to get system delay by (8) formula:

$\left[\begin{array}{c}{d}_1\\ {\mathrm{<figure-callout\; id="2"\; label="d"\; filenames="HDA00001939833100012.png,HDA00001939833100021.png"\; state="\{\{state\}\}">d</figure-callout>}}_2\end{array}\right]=\left[\begin{array}{c}1\\ 1\end{array}\right]$

3, single-loop detector performance computing filter output matrix single filter output

performance index calculation

Applied time series analysis, the performance index that can get each filtering output according to (10) formula are following:

$\left[\begin{array}{c}{\mathrm{\&eta;}}_1\\ {\mathrm{\&eta;}}_2\end{array}\right]=\left[\begin{array}{c}\underset{j=0}{\overset{d_1-1}{\mathrm{\&Sigma;}}}{f}_{1j}/\underset{j=0}{\overset{\&infin;}{\mathrm{\&Sigma;}}}{f}_{1j}\\ \underset{j=0}{\overset{d_2-1}{\mathrm{\&Sigma;}}}{f}_{2j}/\underset{j=0}{\overset{\&infin;}{\mathrm{\&Sigma;}}}{f}_{2j}\end{array}\right]=\left[\begin{array}{c}0.9678\\ 0.9777\end{array}\right]$

4, the index weighter is calculated single filtering output

weighting coefficient by (10) formula

$\left[\begin{array}{c}{\mathrm{\&omega;}}_1\\ {\mathrm{\&omega;}}_2\end{array}\right]=\left[\begin{array}{c}{\mathrm{\&lambda;}}_1/({\mathrm{\&lambda;}}_1+{\mathrm{\&lambda;}}_2)\\ {\mathrm{\&lambda;}}_2/({\mathrm{\&lambda;}}_1+{\mathrm{\&lambda;}}_2)\end{array}\right]=\left[\begin{array}{c}0.9978\\ 0.0022\end{array}\right]$

Combination property detected value by (12) formula control system:

η＝ω ₁η ₁+ω ₂η ₂＝0.9678

The detected value of performance at random that can see system thus is 0.9678, explains that the control performance of this system is better.In fact this design of Controller is at K ₁₂The optimal controller of design in=0 o'clock is at K ₁₂=1 performance decreases but its performance is still better.

The above; Be merely the preferable embodiment of the present invention, but protection scope of the present invention is not limited thereto, any technician who is familiar with the present technique field is in the technical scope that the present invention discloses; The variation that can expect easily or replacement all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection domain of claim.

Claims (6)

1. the control performance pick-up unit of a multiple-input, multiple-output control system is characterized in that this device comprises the decoupling zero wave filter, postpones estimator, single loop property detector and index weighter;

Said decoupling zero wave filter is connected with the index weighter with delay estimator, single loop property detector respectively; Postponing estimator is connected with the single loop property detector; The single loop property detector is connected with the index weighter;

The linear decoupling zero of said decoupling zero wave filter realization system output; Postpone the delay that estimator is used for calculation of filtered output; The single loop performance index of single loop property detector calculation of filtered output; The Comprehensive Control performance index that the index weighter is used to calculate the weighting coefficient of each single loop index and comprehensively obtains mimo systems.

2. control performance detection method of utilizing the said device of claim 1 is characterized in that this method may further comprise the steps:

Step 1: said decoupling zero wave filter is gathered and filtering system's output vector of multiple-input, multiple-output control system, obtains normalized vector, further obtains the eigenwert and the unit character vector of the correlation matrix of normalized vector; Obtain the filtering output matrix at last;

Step 2: the system delay that utilizes said delay estimator calculation of filtered output matrix;

Step 3: the performance index of utilizing said single loop property detector calculation of filtered output matrix;

Step 4: utilize the weighting coefficient of said index weighter calculation of filtered output matrix, the combination property detected value of controlled system.

3. a kind of control performance detection method according to claim 2 is characterized in that the computing formula of said filtering output matrix is:

$\tilde{Y}=\mathrm{MY}$

Wherein:

is the filtering output matrix;

M is the matrix that the unit character vector is formed;

Y is system's output vector.

4. a kind of control performance detection method according to claim 2 is characterized in that the computing formula of said performance index is:

${\mathrm{\&eta;}}_i=\frac{{\left({\mathrm{\&delta;}}_{{\tilde{Y}}_i}^2\right)}_{\mathrm{mv}}}{{\mathrm{\&delta;}}_{{\tilde{Y}}_i}^2}=\frac{\underset{j=0}{\overset{d_i-1}{\mathrm{\&Sigma;}}}{f}_{\mathrm{ij}}}{\underset{j=0}{\overset{\&infin;}{\mathrm{\&Sigma;}}}{f}_{\mathrm{ij}}}$

Wherein:

η _iBe the filtering output matrix

I element

Performance index;

is the filter output matrix i-th element

The minimum variance;

is the filter output matrix

i-th element

of the variance;

d _iBe i element

The system in loop delays;

f _IjFor its noise to i element

Set up the coefficient of AR model.

5. a kind of control performance detection method according to claim 4 is characterized in that the computing formula of said weighting coefficient is:

${\mathrm{\&omega;}}_i=\frac{{\mathrm{\&lambda;}}_i}{\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&lambda;}}_i}$

Wherein:

ω _iBe weighting coefficient;

λ _iI eigenwert for correlation matrix.

6. a kind of control performance detection method according to claim 5 is characterized in that the computing formula of said combination property detected value is:

$\mathrm{\&eta;}=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i{\mathrm{\&eta;}}_i$

Wherein:

η is the combination property detected value.

Images