Stable algorithm of nonparametric identification in case of anomalous measurements

Author(s):  
Юрий Евгеньевич Воскобойников ◽  
Василиса Андреевна Боева

Математические модели многих технических систем имеют вид интегрального уравнения Вольтерра I рода с разностным ядром. Для таких систем задача идентификации заключается в построении оценки для импульсной переходной функции системы по измеренным (с шумами) значениям входного и выходного сигналов и является некорректно поставленной. В недавней работе авторов предложен устойчивый алгоритм идентификации, использующий аппарат сглаживающих кубических сплайнов для вычисления первых производных входного и выходного сигналов. К сожалению, сглаживающие кубические сплайны неудовлетворительно фильтруют аномальные измерения. Поэтому предложен двухшаговый алгоритм идентификации, на первом шаге которого аномальные измерения удаляются с использованием пространственно-локального фильтра, а затем строятся сглаживающие сплайны Volterra integral equation of the first kind often represents stationary dynamic systems. For such a model, the non-parametric identification problem reduces to the estimation of pulse transition characteristics (that is the kernel of integral equation) from the registered noise-contaminated values of input and output signals. To formulate stable solution for identification problem authors propose algorithm that estimates pulse transition characteristics by solving Volterra integral equation of the second kind and involving first derivatives of input and output signals application that corresponds to non-stable problem. Smoothing cubic splines employed in robust calculation of first derivatives allow finding a stable solution of identification problem even when input and output signals of system identified are essentially noise-contaminated. Unfortunately, measured values of input and output signals also contain anomalous measurements such as pulse noises, glitches, etc. Such measurements are poorly smoothable by splines that cause high levels of first derivatives errors and, conversely, significant pulse transition characteristics identification errors of dynamic system. For all the reasons aforementioned, in this paper authors present the new stable two-step identification algorithm in case of anomalous measurements. The first step of the algorithm is for non-linear local-spatial combined filtration procedure of input and output signals that helps to effectively remove anomalous measurements. At the second step, smoothing cubic splines are used to calculate stable first derivatives of previously filtered signals. An extensive computational experiment showed the effectiveness of the proposed algorithm, which allows solving the identification problem with acceptable accuracy in practice even at high intensity of anomalous measurements. The experimental results give reason to recommend this algorithm for solving practical problems of identifying stationary systems, the mathematical model of which is the Voltaire integral equation of the first kind with a difference kernel

Author(s):  
Vasilisa Boeva ◽  
◽  
Yuri Voskoboinikov ◽  
Rustam Mansurov ◽  
◽  
...  

The thermal control system “Heater-Fan-Room” is represented by three different-type interconnected simpler subsystems. In this paper, a “black-box” whose structure is not specified is used as a mathematical model of the system and subsystems due to complexity of physical processes proceeding in these subsystems. For stationary linear systems, the connection between an input and an output of the “black-box” is defined by the Volterra integral equation of the first kind with an undetermined difference kernel also known as impulse response in the automatic control theory. In such a case, it is necessary to evaluate an unknown impulse response to use the “black-box” model and formulate all subsystems and the system as a whole. This condition complicates significantly the solution search of non-parametric identification problems in the system because an output of one subsystem is an input of another subsystem, so active identification schemes are unappropriated. Formally, an impulse response evaluation is a solution of the integral equation of the first kind for its kernel by registered noise-contaminated discrete input and output values. This problem is ill-posed because of the possible solution instability (impulse response evaluation in this case) relative to measurement noises in initial data. To find a unique stable solution regularizing algorithms are used, but the specificity of the impulse response identification experiment in the “Heater-Fan-Room” system do not allow applying computational methods of these algorithms (a system of linear equations or discrete Fourier transformation). In this paper, the authors propose two specific identification algorithms for complex technical systems. In these algorithms, impulse responses are evaluated using first derivatives of identified system signals that are stably calculated by smoothing cubic splines with an original smoothing parameter algorithm. The results of the complex “Heater-Fan-Room” system modeling and identification prove the efficiency of the algorithms proposed. Acknowledgments: The reported study was funded by RFBR, project number 20-38-90041.


Author(s):  
Yuri Voskoboynikov ◽  
◽  
Vasilisa Boeva ◽  
◽  

In a practice, it often happens that complex engineering systems consist of several interconnected different-type simpler subsystems. An adequate model formulation for every subsystem is impractical due to the complexity of physical processes proceeding in the subsystem. In such cases, a non-detailed black-box model is commonly used. For stationary linear systems (or subsystems), the connection between an input and an output of the black-box is defined by the Volterra integral equation of the first kind with an undetermined difference kernel also known as an impulse response in the automatic control theory. It is necessary to evaluate the unknown impulse response to use the black-box model .This statement is a non-parametric identification problem. For complex systems, the problem needs to be solved both for a whole system and for every isolated subsystem that makes identification substantially complex. Formally, impulse response evaluation is a solution of the integral equation of the first kind for its kernel over registered noise-contaminated discrete input and output values. This problem is ill-posed because of possible solution instability regarding measurement noises in initial data. To find a unique stable solution regularizing algorithms are used, but specific input and output signals in impulse response identification experiments do not allow applying computational methods of these algorithms (system of linear equations or discrete Fourier transformation). In this paper, the authors propose two specific-considering identification algorithms for complex engineering systems. In these algorithms, smoothing cubic splines are used for stable calculation of first derivatives of identified system signals. The results of the complex “Heater-Blower-Room” system identification prove the efficiency of algorithms proposed.


2021 ◽  
Vol 10 (4) ◽  
pp. 2245-2253
Author(s):  
C. P. Pandey ◽  
P. Phukan ◽  
K. Moungkang

The integral equations of the first kind arise in many areas of science and engineering fields such as image processing and electromagnetic theory. The wavelet transform technique to solve integral equation allows the creation of very fast algorithms when compared with known algorithms. Various wavelet methods are used to solve certain type of integral equations. To find the most accurate and stable solution of the integral equation Bessel wavelet is the appropriate method. To study the properties of solution of integral equations on distribution spaces Bessel wavelet transform is also used. In this paper, we accomplished the concept of Hankel convolution and continuous Bessel wavelet transform to solve certain types of integral equations (Volterra integral equation of first kind, Volterra integral equation of second kind and Abel integral equation). Also distributional wavelet transform and generalized convolution will be applied to find the solution of certain Integral equations.


2018 ◽  
Vol 19 (6) ◽  
pp. 375-379
Author(s):  
Leszek Cedro ◽  
Krzysztof Wieczorkowski

The paper presents an example of solving the parameter identification problem in case of platform with one degrees of freedom has been also presented. The parameter identification algorithm based on linear parameterization of the platform model and the least square criteria is developed. The desired derivatives of measured signals are estimated by means of designed differentiation filters. The required derivative order depends on the order of differential equations describing the object. The model was identified and verified using measurement results obtained for a real system.


2011 ◽  
Vol 243-249 ◽  
pp. 5403-5407 ◽  
Author(s):  
Ying Lei ◽  
Yan Wu

In this paper, a technique is proposed for non-parametric identification of structural nonlinearity with limited input and output measurements. The identification algorithm is based on the classical Kalman estimator for the displacement and the velocity responses and the recursive least square estimation for the unmeasured excitation and the restoring force. Two different models are used to simulate nonlinear structures: One is a 4-storey shear-frame structure with excitation on the top floor and the nonlinearity occurs at the bottom floor. The other is also a 4-storey shear-frame structure with both excitation and the nonlinearity at the top floor. Two numerical examples are carried out for the two kinds of models. Bouc-Wen hysteretic models are used to simulate the nonlinear impact. The simulation results demonstrate the efficiency of the proposed technique with limited output measurements.


2012 ◽  
Vol 461 ◽  
pp. 686-689
Author(s):  
Li Juan Cao ◽  
Zi Chang Shangguan ◽  
Shou Ju Li

Model identification of dynamic systems in the vibration engineering field has been followed with interest in recent years. A number of identification techniques on this topic are now available, such as parametric or non-parametric identification methods, time domain or frequency domain estimation approaches, etc. The identification approach of nonlinear constitutive model from input-output measurements is proposed. The inverse problem of material characterization is formulated as parameter identification problem that is solved by using optimization procedure. A set of parameters corresponding to the material property can be determined by minimizing objective function which accounts for experimental data and calculated responses of the mechanical model. The performances of the proposed identification approach were evaluated with simulating data. The effectiveness of identification approach is validated by numerical simulation. The investigation results show that the proposed identification algorithm poses good robustness and high identification precision.


Sign in / Sign up

Export Citation Format

Share Document