Wyniki wyszukiwania - BazTech

Ograniczanie wyników

Powiadomienia systemowe

Sesja wygasła!

Znaleziono wyników: 2

Liczba wyników na stronie

Wyniki wyszukiwania

Wyszukiwano:
w słowach kluczowych: równoległe obliczenia

Sortuj według:

Ogranicz wyniki do:

A parallel improved PSO algorithm with genetic operators for 2D inversion of resistivity data

Abril Jorge L., Vasconcelos Marcos A., Barboza Francisco M., Mojica Oscar F.

Acta Geophysica

2022

Vol. 70, no 3

1137--1154

In this paper, an improved particle swarm optimization technique known as elitist-mutated particle swarm optimization (EMPSO) was applied in the 2D electrical resistivity imaging, a complex and highly nonlinear optimization problem. The EMPSO enables better exploration of the search space, by replacing particles with a worse performance by the best particle of the swarm mutated in random positions. Nevertheless, this technique, as any other based on a population of models, costs much computation time in solving optimization problems with a large number of unknown parameters. We addressed this problem by developing a parallel version of the EMPSO that supports pure MPI and hybrid MPI-OpenMP modes, and we named as parallel elitist-mutated PSO (PEMPSO). The solution to the inverse problem is based on minimizing an objective function with a regularization term to create a mathematically stable solution. Total variation and global smoothness regularizations were used in the inversion of synthetic data obtained from simple models and a set of real data of a highly complex geological/geotechnical nature. By virtue of the features of the synthetic models and the geology of the local where the data were acquired, the inversions with total variation regularization provided the best outcomes. Additionally, we have improved the execution time significantly with our parallel solution (the pure MPI model turned out to be better than the hybrid model) in comparison with the sequential version. Cumulative frequency distribution of errors between modeled and observed apparent resistivity data for all experiments was used to validate the PEMPSO technique for estimating resistivity.

Implementation of parallel algorithm of computations of dynamic processes in cluster systems

Byczkowska-Lipińska L., Stakhiv P., Cegielski M.

Przegląd Elektrotechniczny

2010

R. 86, nr 1

20-22

A parallel process of computation of dynamic processes in the electric circuits is based on a diakoptic theory of the portioning of the circuit into the sub-circuits, that are computed independent of each other, in the fixed time intervals. Every process is defined by the complexity of computation. This complexity is changed in the time of process of computations. These properties are utilized to implement computation in parallel systems. Currently the growth of parallel systems has been observed for several years. In this article, idea of cluster system and parallel algorithms are presented. The results of computation on tested parallel system are shown and interpreted.

Równoległa realizacja obliczeń analizy stanów dynamicznych obwodów elektrycznych bazuje na diakoptycznym podziale obwodu na pod-obwody, które w określonych krokach czasowych obliczane są niezależnie od siebie. Każdy taki proces charakteryzuje się dużą złożonością obliczeniową, która w trakcie procesu obliczeniowego może ulegać zmianie. Właściwości te pozwalają na zastosowanie do obliczeń systemów równoległych, których dynamiczny rozwój jest obserwowany od kilku lat. W artykule przedstawiono krótką charakterystykę systemów klastrowych i zastosowanych w nich równoległych algorytmach. Zaprezentowano także eksperymentalne wyniki obliczeń przeprowadzonych na testowanym systemie i dokonano ich interpretacji.