Wyniki wyszukiwania - BazTech

1

Frequency-weighted model order reduction combined with the model decomposition

Raczyński D., Stanisławski W.

Journal of Theoretical and Applied Computer Science

|

2017

|

Vol. 11, nr 1

3--12

EN

Physical objects, in practice, are frequently represented by the LTI (Linear Time Invariant) models consisting of linear differential equations with constant coefficients. The LTI models of complex objects often have a large size. Designing a control system or performing a realtime object simulation on large-scale models may not be possible. Model Order Reduction (MOR) is an operation performed to reduce the size of a model with retaining its most important features. In the case where the reduced model has to approximate the frequency characteristics of the original model within the given range of adequacy, the Frequency-Weighted (FW) method should be used. The FW method requires to select the appropriate parameters of the input and / or output filters in order to obtain the best results. This selection is not an unique operation - which means that it is advisable to use an evolutionary algorithm that performs multiple reduction operations to match the relevant parameters. It is possible to reduce the computational complexity of the FW reduction process by decomposing the model into parts. The purpose of the paper is to compare the Frequency-Weighted order reduction combined with a various model decompositions based on the Schur complement, Schur and the Schur-Sylvester methods. The research is conducted on linearized models of the one-phase zone of the evaporator of the once-through steam BP-1150 boiler. The stability and accuracy of the reduced models are analyzed.

2

A Comparison of Methods for Calculation of Transformation Matrices for Model Order Reduction Using GPU Parallel Computing

Raczyński D.

Zeszyty Naukowe. Elektryka / Politechnika Opolska

|

2013

|

z. 69

91--92

EN

The purpose of this paper is to compare execution times for developed programs for determining transformation matrices in model order reduction using the balanced realization method. Six popular methods are implemented, named the: RPR, SR, BFSR, EIG-SR, EIGBFSR and Obinata-Anderson method. Each algorithm is prepared in two versions, one for execution on the main processor (CPU), and the other one for execution on graphics processor (GPU).

3

Metoda poprawy precyzji algorytmów wyznaczających gramiany sterowalności i obserwowalności

Raczyński D.

Zeszyty Naukowe. Elektryka / Politechnika Opolska

|

2012

|

z. 67

105-109

PL

Celem artykułu jest przedstawienie opracowanej metody poprawy precyzji algorytmów wyznaczających gramiany sterowalności i obserwowalności ciągłych stacjonarnych układów liniowych. W artykule porównano jakość opracowanych algorytmów dla metod: Bartels-Stewarta, Smitha, ADI oraz funkcji znaku.

EN

The article presents the developed methods to improve the precision of algorithms for designating the controllability and observability gramians of LTI systems. The article compares the quality of the developed algorithms for the methods: Bartels-Stewart, Smith, ADI and Sign Function.

4

Cholesky decomposition of controllability and observability gramians using parallel computation with CPU

Raczyński D.

Zeszyty Naukowe. Elektryka / Politechnika Opolska

|

2012

|

z. 68

59--60

EN

Tlic purpose of this paper is to present the advantages of parallel computing the Cholesky decomposition of controllability and observabilily gramians by nieans of NVIDIA general purpose Graphics Processing Unit (GPU). The paper contains results of the conducted examinations comparing speed of the sign function and Smith methods.

5

Controllability and observability gramians parallel computation using GPU

Raczyński D, Stanisławski W.

Journal of Theoretical and Applied Computer Science

|

2012

|

Vol. 6, nr 1

47-66

EN

Algorithms and parallel programs for the controllability and observability gramians computation of the Linear Time Invariant (LTI) systems using Lyapunov equation with an application of the NVIDIA general purpose Graphics Processing Unit (GPU), are presented in the paper. Parallel computing of the gramians on the basis of Lyapunov equation is justified for the large scale systems (n>104) due to the computational cost O(n3). The parallel performance of controllability gramians computation using NVIDIA graphics hardware GTX-465 have been compared with the performance obtained for MATLAB environment employing analogous algorithms. They have also been compared with the performance obtained for lyap function provided by MATLAB environment. The values of maximum computing acceleration were up to 20. The computations have been made on the basis of linearized models of the one-phase zone of a once-through boiler obtained with the finite elements method. The orders of the models were being adapted within the range between 30 and 4200.

6

Vector calculations using the x86 processors family

Raczyński D.

Zeszyty Naukowe. Elektryka / Politechnika Opolska

|

2010

|

z. 63

49-50

EN

The purpose of this article is to present basic features of vector extensions introduced in the x86 family of processors. In order to compare the speed of programs using the vector extensions with those using scalar data, a few programs are developed, in particular programs performing operations on graphical BMP files, counting the indicated integral using the rectangles method and generating fractals.