RU2464616C1 - Method of searching for faults in dynamic unit in continuous system - Google Patents

Abstract

FIELD: physics.

SUBSTANCE: number of dynamic units of a controlled system is determined, reaction of a good system is recorded on an interval at control points, and integral estimates of output signals of the system are determined, for which at the moment of transmitting a test or operating signal to the input of the system with nominal characteristics, integration of signals of that system is simultaneously started for each control point with a weight function equal to the arithmetic average value of moduli of derivatives of its signals at the control points, where averaging is carried out on the number of control points.

EFFECT: broader functional capabilities of the method by applying operating diagnosis, higher noise-immunity of the method of diagnosing continuous automatic control systems by improving distinguishability of defects and lower hardware expenses on calculating the weight function.

1 dwg

Description

The invention relates to the field of monitoring and diagnosing automatic control systems and their elements.

There is a method of finding a faulty unit in a dynamic system (Positive decision dated July 12, 2010 on the grant of a patent for an invention according to application No. 2009123999/08 (033242), MKI ⁶ G05B 23/02, 2010).

The disadvantage of this method is that it determines a defect with a depth of only up to the dynamic block and involves the integration of special test signals using an exponential weight function.

The closest technical solution (prototype) is a method for troubleshooting a dynamic unit in a continuous system (RF Patent for the invention No. 2429518 according to the application No. 201028421/08 (040385), MKI ⁶ G05В 23/02, 2011).

The disadvantage of this method is that it involves the integration of special test signals using an exponential weight function and provides the identification of defects with low discrimination, that is, it has low noise immunity.

The technical problem to which this invention is directed is to expand the functionality of the method by applying working diagnostics (without using a test action), increasing the noise immunity of the method for diagnosing continuous automatic control systems by improving the distinguishability of defects and reducing the hardware cost of calculating the weight function. This is achieved by replacing the exponential weight function with a function that is the arithmetic mean of the modules of the time derivatives of the signals of the system with nominal characteristics, the controlled system and models with trial deviations.

, j=1, 2, …, k на интервале

в k контрольных точках, и определяют интегральные оценки выходных сигналов F_{j ном}(d), j=1, …, k системы, для чего в момент подачи тестового или рабочего сигнала на вход системы с номинальными характеристиками одновременно начинают интегрирование сигналов этой системы для каждой из k контрольных точек с весовой функцией, равной среднему арифметическому значению модулей производных ее сигналов в контрольных точках, где усреднение производится по числу контрольных точек. Для этого на первые входы k блоков перемножения подают сигналы системы, на вторые входы блоков перемножения подают среднее арифметическое значение модулей производных по времени сигналов, выходные сигналы k блоков перемножения подают на входы k блоков интегрирования, интегрирование завершают в момент времени T_к, полученные в результате интегрирования оценки выходных сигналов F_{j нoм}(d), j=1, …, k регистрируют, одновременно определяют интегральные оценки сигналов m моделей для каждой из k контрольных точек, полученные в результате пробных отклонений поочередноThe task is achieved in that they fix the number m of parameters of all dynamic blocks of the controlled system, record the reaction of a known-good system

, j = 1, 2, ..., k on the interval

at k control points, and determine the integral estimates of the output signals F _{j nom} (d), j = 1, ..., k of the system, for which, at the time of supplying a test or working signal to the input of a system with nominal characteristics, the integration of the signals of this system for each of k control points with a weight function equal to the arithmetic mean value of the modules of the derivatives of its signals at control points, where averaging is performed over the number of control points. To do this, the system signals are fed to the first inputs of k multiplication blocks, the arithmetic mean of the time-derivative modules of signals is fed to the second inputs of the multiplication blocks, the output signals of the multiplication blocks are fed to the inputs of the integration blocks k, the integration is completed at time T _k , resulting from integrate the estimates of the output signals F _{j nom} (d), j = 1, ..., k, register, simultaneously determine the integral estimates of the signals of m models for each of k control points obtained as a result of test loony alternately

determine diagnostic signs:

at a minimum, the values of the diagnostic symptom determine the faulty parameter.

The essence of the proposed method is as follows.

The method is based on the use of trial deviations of the model parameters of a continuous dynamic system. To obtain diagnostic signs of dynamic elements, integral estimates are used on the time interval T _k at k control points

The weight function in formula (4) in the form of the average value of the modules of the derived signals at the control points carries information about the importance of the time point in terms of the rate of change of signals at all control points. The higher the average rate of change of the signals, the more weight the output signal is integrated. Using a vector interpretation of expression (3), we write it in the following form

и нормированным вектором (единичной длины) отклонений интегральных оценок сигналов модели с элементами

, полученными в результате пробного отклонения i-го параметра.where φ _i (d) is the angle between the normalized vector (unit length vector) of the deviations of the integral estimates of the object signals with elements

and a normalized vector (unit length) of the deviations of the integral estimates of the model signals with elements

obtained as a result of a trial deviation of the i-th parameter.

Thus, the normalized diagnostic sign (3) is the value of the square of the sine of the angle formed in the k-dimensional space (where k is the number of control points) by the normalized test deviation vectors of the integral estimates of the model signals and the integral estimates of the real deformation of the signals of the diagnostic object.

A test deviation of the parameter, minimizing the value of the diagnostic sign (3), indicates the presence of a defect in this parameter. The range of possible values of a diagnostic feature lies in the interval [0, 1].

Thus, the proposed method for troubleshooting a dynamic unit is reduced to the following operations.

1. As a dynamic system, consider a system consisting of randomly connected dynamic elements, the transfer functions of which contain m parameters.

2. Pre-determine the control time T _To ≥T _PP , where T _PP - the transition process of the system. The transient time is estimated for the nominal values of the parameters of the dynamic system.

3. Fix the number of control points k.

4. At the same time, a test signal x (t) (unit step) or a working signal is input to the input of the control system with nominal parameters, to the input of the controlled system, to the inputs of m models with nominal parameters, in each of which test deviations of one parameter are introduced so that The test deviation in the i-th parameter is introduced in the i-th system.

, реакцию контролируемой системы

, реакции моделей с пробным отклонением в i-ом параметре

в k контрольных точках j=1, 2, …, k на интервале

.5. At the same time, the response of the system with nominal characteristics is recorded.

controlled system response

, reactions of models with a trial deviation in the i-th parameter

at k control points j = 1, 2, ..., k on the interval

.

6. At the same time, integral estimates of the output signals F _{j nom} (d), j = 1, ..., k of the system with nominal characteristics, of the controlled system F _j (d), j = 1, ..., k, of models with trial deviations in i- ohm parameter P _ji (d), j = 1, ..., k; i = 1, ..., m (formula 4). To do this, at the time of input signal input, the integration of signals at each of k control points of the system with nominal characteristics of the controlled system, models with test deviations of block parameters with a weight function equal to the arithmetic mean of the derivatives of the signals at the control points, where averaging is performed over the number of control points, for which the output signals of each system are fed to the first inputs of k multiplication units, to the second inputs of the multiplication units are fed the arithmetic value of the modules of the derived signals of the system at the control points, the output signals of the k multiplication blocks are fed to the inputs of the k integration blocks, the integration is completed at time T _k , obtained by integrating the estimates of the output signals F _{j nom} (d), j = 1, ... , k, F _j (d), j = 1, ..., k, P _ji (d), j = 1, ..., k; i = 1, ..., m are recorded.

7. The deviations of the integral estimates of the model signals obtained as a result of test deviations of the parameters of the corresponding blocks are determined

ΔP _ji (d) = P _ji (d) -F _{j nom} (d), j = 1, ..., k; i = 1, ..., m.

8. Determine the normalized deviation values of the integral estimates of the model signals obtained as a result of trial deviations of the corresponding parameters by the formula:

, j=1, …, k; i=1, …, m.

, j = 1, ..., k; i = 1, ..., m.

9. The deviations of the integral estimates of the signals of the controlled system for k control points from the nominal values ΔF _j (d) = F _j (d) -F _{j nom} (d), j = 1, ..., k are determined.

10. Calculate the normalized values of the deviations of the integrated estimates of the signals of the controlled system according to the formula:

, j=1, …, k.

, j = 1, ..., k.

11. Calculate the diagnostic signs of the presence of a faulty parameter by the formula (3).

12. At a minimum, the values of the diagnostic symptom determine the defective parameter.

Since the diagnostic signs (3) have a range of possible values limited by the interval [0, 1], the difference between the closest to the minimum sign and the minimum sign (which indicates a defective parameter) quantitatively characterizes the distinguishability of this defect, taking into account the location of the parameter on the structural diagram, of the form and other parameters of the transfer functions of the blocks and all the diagnostic conditions under which these values of diagnostic signs are obtained (the number and location of control points, the value and Tervala T _k). The best distinguishability of defects is ensured when the indicated difference is equal to unity (in terms of vector interpretation, the normalized strain vectors of integral transformations of the dynamic characteristics of these parameters for trial deviations are orthogonal). The worst distinguishability is when the indicated difference is zero (in terms of vector interpretation, the normalized strain vectors of the integral transforms of the dynamic characteristics of these parameters are collinear for trial deviations).

Consider the implementation of the proposed method for finding a single parametric defect for a system whose structural diagram is shown in the figure.

Transfer functions of blocks:

;

;

,

;

;

,

nominal values of parameters: T ₁ = 5 s (J ₁ ); k ₁ = 1 (J ₂ ); k ₂ = 1 (J ₃ ); T ₂ = 1 s (J ₄ ); k ₃ = 1 (J ₅ ) and T ₃ = 5 s (J ₆ ).

When modeling, we will use a pseudo-random signal as the input signal (when modeling, we used the Band-Limited White Noise block in the Matlab environment). The control time we choose T _to equal 10 s.

The value of the test deviations of the model parameters is chosen equal to 10%.

Simulation of the processes of searching for defects in the form of a decrease in the parameter T ₁ by 20% leads to the calculation of diagnostic features according to the formula (3): J ₁ = 5.793e-005; J ₂ = 0.1457; J ₃ = 0.266; J ₄ = 0.3127; J ₅ = 0.1405; J ₆ = 0.2147. Distinctness of the defect: ΔJ = J ₂ -J ₁ = 0.1457.

For comparison, we present diagnostic signs of the presence of a faulty unit using an exponential weight with one integration parameter α = 0.5 (Positive decision dated 04/11/2011 on the grant of a patent for an invention No. 2429518 according to application No. 2010128421/08 (040385), MKI ⁶ G05B 23 / 02, 2011): J ₁ = 2.293e-008; J ₂ = 1.469e-007; J ₃ = 0.7843; J ₄ = 0.7828; J ₅ = 0.07422; J ₆ = 0.07399. Distinctness of the defect ΔJ = J ₂ -J ₁ = 1.2397e-007.

The above results show that the actual distinguishability of finding defects by this method is higher, therefore, the noise immunity of the method will also be higher.

Simulation of defects search processes in the form of a decrease in the parameter T ₂ by 20% for a given diagnostic object using a differential weight and with the same input signal gives the following values of diagnostic signs: J ₁ = 0.2086; J ₂ = 0.7451; J ₃ = 0.006981; J ₄ = 0.001517; J ₅ = 0.7004; J ₆ = 0.6441.

Distinctness of the defect ΔJ = J ₃ -J ₄ = 0.005464.

For comparison, we present diagnostic signs of a faulty unit using an exponential weight with one integration parameter α = 0.5: J ₁ = 0.7828; J ₂ = 0.7825; J ₃ = 3.695e-006; J ₄ = 1.52e-008; J ₅ = 0.7455; J ₆ = 0.7462. Distinctness of the defect: ΔJ = J ₃ -J ₄ = 3.6798e-006.

Modeling of defects search processes in the form of a decrease in the parameter T ₃ by 20% for a given diagnostic object under the same conditions gives the following values:

J ₁ = 0.2186; J ₂ = 0.1878; J ₃ = 0.5691; J ₄ = 0.6809; J ₅ = 0.003594; J ₆ = 2.192e-005.

Distinctness of the defect: ΔJ = J ₅ -J ₆ = 0.003594.

For comparison, we present the diagnostic signs of a faulty unit with one integration parameter α = 0.5:

J ₁ = 0.07407; J ₂ = 0.07384; J ₃ = 0.7476; J ₄ = 0.7457; J ₅ = 7.831e-007; J ₆ = 1.531e-010.

Distinctness of the defect ΔJ = J ₅ -J ₆ = 7.829469е-007.

The minimum value of the diagnostic sign in all cases correctly indicates a defective unit, and this method in all cases improves the actual distinguishability of defects, therefore, increases the noise immunity of the diagnosis.

In addition, the inventive method allows diagnosing in the conditions of the real functioning of the diagnostic object (working diagnosis).

Claims (1)

1. The way to find a faulty block in a dynamic system, based on the fact that the number m of parameters included in the system is fixed, the monitoring time T _k ≥T _{PP is} determined, the input signal x (t) is used on the interval t∈ [0, T _K ], fix the number k of control points of the system, register the response of the controlled system f _j (t), j = 1, 2, ..., k, record the reaction of the system with nominal characteristics f _{j nom} (t), j = 1, 2, ..., k, on the interval t∈ [0, T _k ] at k control points, determine the integral estimates of the output signals of the system, for which, at the time of the signal at the input of a system with nominal characteristics, the integration of system signals at each of k control points is simultaneously started by applying the output signals of the system to the first inputs of k multiplication blocks, the weight function is fed to the second inputs of the multiplication blocks, the output signals of the multiplication blocks are fed to the inputs of the integration blocks k, the integration is completed at time T _k obtained by integrating the evaluation of the output signals recorded determine integral evaluation of the output signals to a model for each of k control points obtained as a result of test deviations of each of the m parameters of all blocks, for which a test deviation is introduced into each parameter of the dynamic system and integral estimates of the model output signals for the weight function and input signal x (t) obtained as a result are found integrating the estimates of the output signals for each of k of each of the m parameters of all blocks, for which the corresponding test deviation of the i-th parameter of the transfer function of all blocks is dynamically introduced into each i-th model th system and find the integral estimates of the output signals of systems with test deviations for the same test or working signal x (t) obtained by integrating the estimates of the output signals for each of the k control points of each of the m test deviations P _ji (d), j = 1, ..., k; i = 1, ..., m, register, at the same time, the test or operating signal x (t) is fed to the input of the controlled system, the integral estimates of the signals of the controlled system are determined for k control points F _j (d), j = 1, ..., k, obtained the values are recorded, the deviations of the integral estimates of the model signals obtained as a result of the trial deviations of the corresponding parameters of all blocks ΔP _ji (d) = P _ji (d) -F _{j nom} (d), j = 1, ..., k; i = 1, ..., m, determine the normalized deviations of the integral estimates of the model signals obtained as a result of trial deviations of the corresponding parameters of all blocks

determine the deviations of the integral estimates of the signals of the controlled system for k control points ΔFj (d) = F _j (d) -F _{j nom} (d), j = 1, ..., k, determine the normalized values of the deviations of the integral estimates of the signals of the controlled system of control points and each from m test deviations are recorded, deviations of the integral estimates of the model signals obtained as a result of test deviations of the corresponding parameters ΔP _ji (d) = P _ji (d) -F _{j nom} (d), j = 1, ..., k are determined; i = 1, ..., m, determine the normalized deviations of the integral estimates of the model signals obtained as a result of trial deviations of the corresponding parameters from the relation

determine the integral estimates of the signals of the controlled system for k control points F _j (d), j = 1, ..., k, determine the deviations of the integral estimates of the signals of the controlled system for k control points from the nominal values ΔF _j (d) = F _j (d) - F _{j nom} (d), j = 1, ..., k, determine the normalized values of the deviations of the integral estimates of the signals of the controlled system from the relation

determine diagnostic signs from the ratio

at the minimum of the diagnostic sign, a faulty parameter is determined, characterized in that at the same time a test or operating signal x (t) is supplied to the input of the system with nominal characteristics, to the input of the controlled system, to the inputs of m models with nominal characteristics, in each of which a test deviation of one from the parameters so that a trial deviation is introduced into the i-th system in the i-th parameter, as the dynamic characteristics of the system, integral estimates are used obtained for the weight function equal to the arith average eticheskomu modules time derivatives of the output signals of the system in various control points from the relationship

determine the integrated estimates of the output signals F _{j nom} (d), j = 1, ..., k, of the system with nominal characteristics, for which, at the time of the input signal to the input of the system with nominal characteristics, the integration of the system signals in each of k control points at the same time begins weight function, by applying to the first inputs of k blocks of multiplication of the system signals to the second inputs of the blocks of multiplication serves the arithmetic average of the time derivatives of the output signals of the system with nominal characteristics, output e signals k multiplying unit is supplied to the inputs k of integration blocks the integration is completed at time T _k obtained by integrating the evaluation of the output signals F _{j prefecture} (d), j = 1, ..., k are recorded similarly define integral evaluation signal m Models for each of k control points, obtained as a result of test deviations of each of m parameters, obtained by integrating an estimate of the output signals for each of k control points, of each of m test deviations P _ji (d), j = 1, ..., k; i = 1, ..., m, is used to calculate diagnostic features.